Essays on the Development of the American Economy
MASSACHUS9TTS INSTITUTE
OF TECHNOLOGY
by
Richard A. Hornbeck
B.A. Economics
University of Chicago, 2004
JUN 0 8 2009
LIBRARIES
Submitted to the Department of Economics in
Partial Fulfillment of the Requirements for the Degree of
APSCHIVES
Doctor of Philosophy in Economics
at the
Massachusetts Institute of Technology
June 2009
© 2009 Richard A. Hornbeck. All rights reserved.
The author hereby grants to MIT permission to reproduce and to distribute publicly
paper and electronic copies of this thesis document in whole or in part
in any medium now known or hereafter created.
Signature of Author .............................................
Department of Economics
May) 5, 2009
Certified by...............................
/
' Michaelb3reenstone
3M Profe ssor of Environmental Economics
Thesis Supervisor
r,
Certified by....
Esther Duflo
Abdul Latif Jameel Professor of Pov rty Alleviation and Development Economics
Thesis Supervisor
T
Certified by.
V
LDaron Acemoglu
Charles P. Kindleberger Professor of Applied Economics
Thesis Supervisor
Accepted by
Esther Duflo
Abdul Latif Jameel Professor of Poverty Alleviation and Development Economics
Chairman, Departmental Committee on Graduate Studies
Essays on the Development of the American Economy
by
Richard A. Hornbeck
Submitted to the Department of Economics on May 15, 2009
in partial fulfillment of the requirements for the Degree of
Doctor of Philosophy in Economics
ABSTRACT
The first essay analyzes the impact of the 1930's American Dust Bowl and investigates how
much the short-term costs from erosion were mitigated by long-term adjustments. Exploiting
new data collected to identify low, medium, and high erosion counties, estimates indicate that the
Dust Bowl led to substantial immediate decreases in agricultural land values and revenues. Until
at least the 1950's, however, there was limited reallocation of farmland away from activities that
became relatively less productive due to erosion. Relative changes in agricultural land values and
revenues indicate that the annualized long-term cost was 86% of the short-term cost to
agriculture. Substantial out-migration reflects the large cost of the Dust Bowl, and was an
important channel through which short-term costs were partly mitigated.
The second essay examines the impact on agricultural development from the introduction of
barbed wire fencing to the American Plains in the late 19 th century. Farmers were required to
construct fences to be entitled to compensation for damage by others' livestock. From 1880 to
1900, the introduction and universal adoption of barbed wire greatly reduced the cost of fences,
relative to predominant wooden fences, most in counties with the least woodland. Over that
period, counties with the least woodland experienced substantial relative increases in settlement,
land improvement, land values, and the productivity and production share of crops most in need
of protection. This increase in agricultural development appears partly to reflect farmers'
increased ability to protect their land from encroachment. States' inability to protect this full
bundle of property rights on the frontier, beyond providing formal land titles, might have
otherwise restricted agricultural development.
The third essay quantifies agglomeration spillovers by comparing the growth of total factor
productivity (TFP) among incumbent plants in "winning" counties that attracted a large
manufacturing plant to "losing" counties that were the plant's second choice. Five years after the
opening, incumbent plants' TFP is 12% higher in winning counties. This effect is larger for
plants with similar labor and technology pools as the new plant. We find evidence of increased
wages in winning counties, indicating that profits increase by less than productivity.
Thesis Supervisor: Michael Greenstone
Title: 3M Professor of Environmental Economics
Quantifying Long-term Adjustment to Environmental
Change: Evidence from the American Dust Bowl
Richard A. Hornbeck*
Abstract
The first essay analyzes the impact of the 1930's American Dust Bowl and investigates how much the short-term costs from erosion were mitigated by long-term
adjustments. Exploiting new data collected to identify low, medium, and high erosion
counties, estimates indicate that the Dust Bowl led to substantial immediate decreases
in agricultural land values and revenues.
Until at least the 1950's, however, there
was limited reallocation of farmland away from activities that became relatively less
productive due to erosion. Relative changes in agricultural land values and revenues
indicate that the annualized long-term cost was 86% of the short-term cost to agriculture. Substantial out-migration reflects the large cost of the Dust Bowl, and was an
important channel through which short-term costs were partly mitigated.
*I thank my advisors Michael Greenstone, Esther Duflo, Daron Acemoglu, and Peter Temin. I also thank
David Autor, Abhijit Banerjee, Geoff Cunfer, Joe Doyle, Greg Fischer, Price Fishback, Claudia Goldin,
Raymond Guiteras, Gary Libecap, Bob Margo, Ben Olken, Paul Rhode, Claudia Rei, Cate Rosenthal,
Tavneet Suri, Rob Townsend, John Wallis, and Noam Yuchtman for their comments, suggestions, and/or
generosity with data. I am grateful to Lisa Sweeney, Daniel Sheehan, and the GIS Lab at MIT for their
assistance with GIS software; as well as to Christopher Compean, Lillian Fine, Phoebe Holtzman, Paul
Nikandrou, and Praveen Rathinavelu for their valuable research assistance. For supporting research expenses,
I thank the MIT Schultz Fund, the World Economy Lab, and UROP program. hornbeck@mit.edu
One recurrent theme within economics is that production adjusts more to a shock in the
"long-run" than in the "short-run" (e.g., Samuelson's Le Chatelier principle). The short-run
costs from a shock may be mitigated in the long-run by economic agents' adjustments. This
is a central issue in the context of global climate change and local environmental collapse
(Diamond 2005). While short-run costs from weather fluctuations may be large, the implied
effect for a permanent change in climate may be an overestimate if agents are able to adjust
(Deschenes and Greenstone 2007, Dell et al. 2008, Guiteras 2008).
Similar issues motivate work by Blanchard and Katz (1992) and others on regional economic adjustment to labor demand shocks and the resulting geographic allocation of labor.
This relates to a general question of how quickly resources can be reallocated in response to
changes in prices or relative productivity. While such responses may be slow and difficult to
observe in response to modern shocks, historical shocks provide settings in which to identify
economic adjustments that occur over a long period of time (Carrington 1996, Margo 1997,
Duflo 2004).
Empirical challenges arise in isolating the effect of an historical shock from other economic
changes.
The shock must be large and observable to people at the time, but to allow
identification of its effects it must also be differential across areas. Large geographicallydifferentiated shocks are rare, and there is typically not detailed data on what happened in
the immediate aftermath of the shock. This makes it difficult to study the specific adjustment
strategies that might have been adopted in reaction to the shock. This data scarcity explains
why the historical literature has generally focused on long-run changes in population (Davis
and Weinstein 2002, Miguel and Roland 2006, and Redding and Sturm forthcoming).
This paper analyzes the aftermath of tremendous permanent soil erosion during the 1930's
American Dust Bowl. The Dust Bowl erosion was quite damaging, and its effects varied
across counties within a state. Detailed data allow for an examination of many dimensions
of adjustment over more than 60 years. This paper focuses on the speed and magnitude
of adjustment in the agricultural sector, the resulting difference in short-run and long-run
costs, and the reallocation of labor.
The Dust Bowl was a period of extreme soil erosion on the American Plains, unexpectedly
brought about by the combination of severe drought and intensive land-use. Strong winds
swept topsoil from the land in massive dust storms, and occasional heavy rains carved deep
gullies in the land. By the 1940's, many Plains areas had cumulatively lost more than 75%
of their original topsoil.
The paper begins by setting out a model of agricultural production in the short-run
and long-run. The model highlights that the cost of the Dust Bowl will be immediately
capitalized into land values. In the long-run, land allocations adjust toward less-affected
activities. In general equilibrium, adjustment may also occur through out-migration and an
expansion of local industry. Thus, there are multiple potential margins of adjustment, and
this paper examines many of them.
The key to this analysis is the existence of substantial local variation in erosion following
the Dust Bowl. In particular, the empirical analysis is based on a comparison of areas that
became severely, moderately, or lightly eroded. Erosion levels are combined with Census
data to form a balanced panel of counties from 1910 to the 1990's. The regressions compare
changes in more-eroded and less-eroded counties within the same state, after adjusting for
differences in pre-Dust Bowl characteristics.
The Dust Bowl imposed substantial costs on Plains agriculture.
From 1930 to 1940,
immediately following the Dust Bowl, more-eroded counties experienced substantial and
permanent relative declines in agricultural land values.
The per-acre value of farmland
declined by 28% in high erosion counties and 16% in medium erosion counties, relative to
changes in low erosion counties. Assuming that the low erosion counties were not affected,
the decline in land values indicates a total capitalized economic loss from the Dust Bowl of
$153 million in 1930 dollars ($1.9 billion in 2007 dollars). Given average land values in 1930,
this is the total value of farmland equal to an area roughly the size of Oklahoma.
More-eroded counties also experienced substantial immediate declines in agricultural
revenues per-acre of farmland. Under strong theoretical assumptions, the relative percent
changes in agricultural land values and revenues imply that the annualized long-term cost,
in present discounted value, was 86% of the short-term cost to agriculture. That is, for each
1% that revenue declined from 1930 to 1940, land values declined by 0.86%. Similar levels
of cost persistence are implied by the estimated path of recovery in revenues.
One potential margin of adjustment to the Dust Bowl was in land-use decisions, but
there was limited and slow adjustment. Total farming experienced little decline, consistent
with an inelastic demand for land in non-agricultural sectors. Within the agricultural sector,
adjustment appears to have been possible: there was a decrease in the relative productivity
of land for crops (compared to animals) and for wheat (compared to hay). Until at least the
1950's, however, there was little adjustment of land from crops to pasture or from wheat to
hay.
The paper then examines why land-use adjustment was slow. Credit constraints may
have been important, and adjustment was greater in the more-eroded areas that had more
banks prior to the 1930's, compared to those with few banks. Evidence is mixed on the role
of tenant farming in slowing land-use adjustment. Government programs do not appear to
have discouraged adjustment, at least not in more-eroded counties relative to less-eroded
counties, as payments were not targeted toward more-eroded counties.
A second potential margin of adjustment was labor, through migration or movement between sectors. Consistent with contemporary and historical accounts, there was a substantial
population decrease from 1930 to 1940 in more-eroded counties relative to less-eroded counties, at the same time as substantial out-migration from the entire region. By 1940, unemployment was slightly higher and proxies for wages were lower. Equilibrium was reestablished
through further declines in population, rather than an increase in local industry.
One caveat of examining within-region variation from a shock is the inability to observe
aggregate adjustment through technological innovation. Instead, this paper analyzes changes
in technology choice, treating the technological possibility frontier as common to all areas.
Overall, this paper finds that the Dust Bowl imposed substantial costs on Plains agriculture. Immediate adjustment took place mainly through decreased land values and population, while agricultural land-use only adjusted slowly. There was limited mitigation of
short-run agricultural losses. Migration reflected these large losses, and was an important
channel through which economy-wide consequences were mitigated, both in the short-run
and long-run.
The paper is organized as follows. Section I provides historical background on the Dust
Bowl and aggregate trends over the
2 0 th
century. Section II models the short-run and long-
run response of agricultural production to the Dust Bowl, and discusses how population,
industry, and other locations might also change. Section III describes the measurement of
erosion levels and presents baseline summary statistics by erosion level. Section IV outlines
the empirical methodology, Section V presents the results, and Section VI concludes.
Historical Background
I
I.A
The Dust Bowl and Massive Soil Erosion: 1931 to 1938
Since the late
19 th
century, agricultural activity had increased substantially on the Plains.
World War I then increased demand for American agricultural goods, leading to high prices
and a boom period in Plains agriculture.
Farming continued to expand and European
production recovered after WW1, contributing to low prices for agricultural goods in the
1920's. Thus, despite high yields, farms were struggling even prior to the Great Depression.
By the time moderate drought began on the Plains in 1931, farms had increasingly
plowed up native grasslands.
Dry conditions persisted, contributing to a loss in ground
cover that made the land susceptible to self-perpetuating dust storms (wind erosion). The
hardened bare ground also experienced substantial runoff during the occasional heavy rain
storm (water erosion).
In 1934 and 1936, there were devastating droughts and widespread crop failure. From
1934 to 1936, especially large dust storms blew topsoil across the Plains. The occasional fierce
rain, especially in 1935, carved deep gullies in farmland. This period of massive erosion (the
Dust Bowl period) is considered to have ended by 1941 at the latest, along with the US entry
into WW2, the end of the Depression, and the return of generally wetter weather.
There is limited precise information, however, on when erosion occurred in particular
areas, as the stock of topsoil was not recorded in comparable units at different points in
time. The Soil Erosion Service (SES), established in 1933 and expanded to become the Soil
Conservation Service (SCS) in 1935, published erosion maps from 1933 to December 1937, at
which point there is a gap in the National Archives collection of soil erosion maps until the late
1940's. The SCS marked a very general boundary for which areas experienced severe wind
erosion from 1935 to 1936, in 1938, and from 1935 to 1938. There are few historical accounts
or references to severe erosion occurring after 1938, despite some incorrect information that
has circulated. 1 Thus, I consider the Dust Bowl erosion to have occurred between 1931 and
1938, i.e., between the collection of census data in 1930 and 1940.
1
Worster (1979, p.30) presents a map of the SCS wind erosion regions (1935-1936, 1938, 1935-1938) that
also adds a substantial wind erosion area in 1940. Geoff Cunfer discovered the source of this error: a 1940
USDA document, which cites a December 8, 1939 Washington Evening Star newspaper article, which in
turn cites information provided by SCS technicians. The displayed 1940 wind erosion region was projected
in 1939, and the 1940 USDA document states that these projections proved incorrect and that there turned
out to be little blowing. The 1939 newspaper article map also displays how the Dust Bowl region evolved
from 1935-1939. In addition to the common 1935-1936 and 1938 regions, this map shows a 1937 region that
is similar to and contained within the 1935-1936 region and a 1939 region that is at the center of the 1938
region and is much smaller. Neither Cunfer nor I has found these 1937 and 1939 regions in other archived
publications. My interpretation is that these regions were included by the newspaper to fill in the time series
from 1935-1939, rather than as a reflection of particular new blowing in those years.
I.B
Causes of the Dust Bowl and Government Policy
The Dust Bowl was caused by a combination of prolonged severe drought and intensive
land-use, though it is debated how much weight should be attributed to each factor.2 New
Deal government officials emphasized that an unsustainable amount of Plains land had been
plowed for crops and that the remaining grassland was over-grazed by livestock.
There
was a fear that the region would become the once-imagined "American Desert" if private
land-use practices were allowed to continue (Science 1934, News-week 1936). Attributing
the Dust Bowl to unsustainable land-use became an important rationalization for New Deal
agricultural policies.
One explanation for the Depression was farmers' lack of purchasing power, due to low
prices for agricultural goods. The 1933 Agricultural Adjustment Act was designed to raise
prices by paying farmers to reduce planted acreages. The government also purchased livestock at above-market prices, canning some meat and destroying the rest (Leuchtenburg
1963, Saloutos 1982).
As these initial policies became less popular and the Supreme Court declared aspects unconstitutional, the policies' stated purpose shifted toward conservation. The 1935 and 1936
Soil Conservation and Domestic Allotment Acts were partly an effort to combat erosion and
preserve the country's farmland resources, but they also represented the new banner under
which to reduce agricultural production (Rasmussen and Baker 1979, Phillips 2007). The
1938 Agricultural Adjustment Act combined efforts to maintain higher prices with this new
focus on erosion prevention, nominally making a portion of acreage reduction payments conditional on conservationist land-use adjustments. The SCS established soil erosion control
projects, aimed at demonstrating the effectiveness of their recommended soil conservation
techniques (SCS 1937). These adjustments were mainly plowing along contour lines, retaining crop residues to provide ground cover, fallowing land with protective cover, planting
2
The typical historical view emphasizes the over-exploitation of land (SCS 1955, Worster 1979, Hurt 1981),
as did most contemporaries (SCS 1935 and 1939, Hoyt 1936, Wallace 1938). Cunfer (2005) emphasizes the
impact of severe weather.
alternate strips of cash crops and drought-resistant crops, or converting cropland back to
grassland. 3
I.C
Responses to the Dust Bowl
The Dust Bowl is associated with substantial out-migration, most famously described in
Steinbeck's The Grapes of Wrath. There is little empirical evidence, however, that separates Dust Bowl migration from other sources of change in Plains states' population. All
migrants from the region may have been inappropriately (and derogatorily) lumped together
as "Okies." 4
Most historical discussions of agricultural land-use have focused on its role in causing the
Dust Bowl, but many give the impression that there was substantial adjustment following
the Dust Bowl. Historical accounts of government programs generally suggest that these
encouraged substantial land-use adjustment.5
Hansen and Libecap (2004) present evidence that externalities played an important role
in causing the Dust Bowl. Farmers could discourage wind erosion by fallowing land or
converting land to grasslands and pasture, but much of the benefit would be captured by
neighboring farms. They argue that increased farm sizes and soil conservation districts
lowered these coordination problems, and that resulting production adjustments prevented
the Dust Bowl's reoccurrence. 6
Agricultural adjustment to the Dust Bowl would be expected to reflect the substantial
decrease in productivity due to soil erosion. In particular, erosion may have changed the
3
The SCS also took greater responsibility over the administration of Civilian Conservation Corps work
camps, which employed workers on public projects to reduce unemployment. Some of this work focused on
preventing erosion: terracing land, rehabilitating gullies, and planting trees as wind breaks.
4
Worster (1979, p51) describes interviews with migrants registering at Federal Emergency Relief offices
around the country: only 12% of families from Oklahoma attributed their migration to farm failure, 17% of
families from Kansas, and 16% of families from Colorado.
5
For example, see Hurt (1985). In contrast, Cunfer (2005) notes that overall land-use patterns appear
surprisingly stable from 1920 to 2000. Yet Cunfer's aggregate analysis of land-use may overlook important
adjustments within agricultural production (Hansen 2005).
6
This is in line with much of the historical literature However, Worster (1979) argues that the Dust
Bowl did reoccur to some degree, and that conservation never replaced capitalism as the dominant farming
mentality.
relative returns to different productive activities. If the production of crops is more sensitive
to soil quality, then farmers may shift more-eroded land into pasture for animals. Similarly, land may be shifted from soil-sensitive crops (wheat) to more erosion-resistant crops
(hay/grass). Agricultural adjustment will also interact with migration and adjustment in
non-agricultural sectors, as discussed later.
Aggregate Trends in Agriculture and Population on the Plains
I.D
Figure 1 presents aggregate changes in agriculture and population on the Plains over the
2 0 th
century. 7 Panels A and B show changes in the per-acre value of farmland and farm
revenue in logs.8 Average farmland became less valuable and less productive from 1910 to
the 1940's. This partly reflects an increase in total farmland (Panel C), which expanded into
less valuable areas. Average land values and revenues rose after total farmland stabilized in
the 1940's.
Panel D shows that the total amount of cropland, as a fraction of total farmland, was
constant from 1925 to 1930, decreased from 1930 to 1945, and was then mostly constant.
This decrease is consistent with accounts that Plains farming had become too intensive prior
to the 1930's and then found it necessary to scale back cultivation. However, this decrease
is also consistent with an expansion of farming from 1930 to 1945 into areas less-suited for
growing crops.
Panel E shows that there was a substantial increase from 1930 to 1940 in the proportion
of cropland kept fallow. Much of this increase was short-lived, but fallowing remained above
its level in 1930. Panel F shows that there was a gradual increase in average farm sizes after
7
Aggregate numbers are calculated by summing over the same 769 Plains counties in every reported
period. The only exception is in Panel E. Data on land fallowing are available from 1925 to 1945 for all 769
counties, when fallow land is defined as cropland left idle or in cultivated summer fallow. In 1950 and after,
data are only available in 587 counties for land in cultivated summer fallow. Following Hansen and Libecap
(2004), this series is treated as comparable and the sample is restricted to the 587 counties over the entire
period.
8
The value of farmland is divided by the CPI, as land is an asset whose value will change with the price
of consumption. The value of revenue is divided by the PPI for all farm products, to focus on changes in
productivity rather than output prices.
1930. The aggregate changes in panels E and F are consistent with results presented by
Hansen and Libecap (2004): increased farm sizes may have allowed farmers to internalize
the positive externalities from preventing wind erosion by fallowing land and decreasing
cultivation shares.
Panel G shows that total population grew at a roughly constant rate from 1910 to 2000,
with somewhat slower growth from 1930 to 1950.
Note that, from 1930 to 1940, total
population declined in the five central Plains states (Oklahoma, Kansas, Nebraska, South
Dakota, North Dakota) between 3% and 7%. Panel F shows that the fraction of population
in rural areas declined at a constant rate from 1910 to 1940, and a somewhat quicker rate
from 1940 to 1970. Some of this decline is mechanical: as population grows, more people
will be in non-rural areas (places with more than 2,500 inhabitants).
Figure 1 provides an overall empirical context in which to interpret the estimated relative changes. However, it is difficult to isolate the effect of the Dust Bowl in these aggregate
changes.
Even relative to other areas of the country, other shocks may substantially af-
fect these outcomes (e.g., WW1, the Great Depression, WW2, technological change, price
changes). The empirical analysis that follows will attempt to isolate the effects of the Dust
Bowl, controlling for state-specific changes over each time period.
II
Theoretical Framework
This section outlines a model in which permanent erosion during the Dust Bowl may affect
agricultural production differently in the "short-run" and "long-run."
In the short-run,
farmers can only adjust variable inputs. In the long-run, farmers can adjust product choices
or fixed inputs at some cost. This model helps to interpret the speed and magnitude of
adjustment in agricultural production decisions, and to compare short-run and long-run
costs using observed changes in land values, revenue, and inputs.
The effects on agriculture are then interpreted within a basic model of long-run supply
and demand across locations, sectors, and factor inputs. This model guides the empirical
analysis of migration and non-agricultural outcomes, and informs how cross-county spillovers
influence the estimated relative effect of the Dust Bowl within states.
II.A
A Model of Agricultural Production
Assume perfectly competitive markets, in which a farmer chooses production decisions in
every period to maximize the present discounted value of profits. The farmer produces one
good and allocates a share 0 of land between two production technologies, FI(0, V) and
F 2 (1- 0, V2 ), that are each a function of variable input choices V. The variable inputs reflect
factors that can be adjusted quickly, such as labor, and are purchased in each period at a
constant price. 9 The two technologies reflect different production methods or choices of fixed
inputs that can only be adjusted slowly, such as land characteristics (cropland vs. pasture)
or product choices (wheat vs. hay).
For any given land allocation, the farmer chooses variable inputs to maximize profits.
Define It1(0) and II2(1 - 0) as the maximum obtainable profit after allocating land to each
technology. Given these maximum obtainable profits, the farmer chooses 0 to maximize total
profits:
I (0) + I2(1-
0). Assuming that these two functions are differentiable and concave,
an optimal interior 0 is chosen such that I1',()
= YI,(
-
).10
The farmer obtains an initial profit i7r' = 11(0) + 12 (1 - 0). The value of land is assumed
to equal the present discounted value of profits,
--
, where 3 is a constant discount factor.
9
This baseline model assumes that variable inputs are supplied in competitive markets, in which agriculture is only a small part. This rules out general equilibrium effects where input prices depend on agricultural
production.
10This analysis will focus on interior solutions, where technological change is simply an adjustment along
an existing production possibility frontier. If the initial equilibrium were a corner solution and only one
technology were used, then subsequent adoption of the other technology could be interpreted as technological
growth.
This assumption is consistent with competition and free entry in land markets, with the
marginal land buyer having access to credit at a fixed interest rate.
II.B
1
Agricultural Production After a Permanent Shock
An unexpected and permanent shock at t = 0 is assumed to decrease the relative profitability
of the first technology. Define periods t < T as the "short-run" when variable inputs can
adjust but the land allocation is constrained at its previous level (0 = 0). The land allocation
is then completely adjusted when t = T, after which is the "long-run." Assume initially that
adjusting the land allocation is costless, but the model will then highlight the impact of
adjustment costs.
This artificial distinction between the "short-run" and "long-run" can be interpreted as
a simplified case of an unconstrained optimal adjustment model in which several factors
could delay adjustment. First, adjustment costs could be convex, e.g., if needed capital or
other inputs have an upward sloping supply curve in each period. Second, adjustment costs
could decline over time due to learning-by-doing or other positive spillovers in technological
adoption (Griliches 1957, Foster and Rosenzweig 1995). Third, the potential for learning to
resolve uncertainty about optimal adjustments could delay investment responses (Dixit and
Pindyck 1994, Bloom et al. 2007). Fourth, the original land allocation could have required
fixed investments that depreciate over time, and would be lost during adjustment (Chari
and Hopenhayn 1991).
Change in Profits. By assumption, the relative profitability of the first technology
decreases in the short-run, given the new optimal choice of variable inputs. For analytical
convenience, assume that it decreases by a constant percentage: Ii (0) decreases to 6~i1 (0),
where 6 E (0,1). The land allocation constraint binds because H',(0) < II2 (1- 0). The
farmer earns a short-run profit
qrsR
= 6I11(0) +
r12 (1
- 0). In the long-run, a new optimal
"1If this assumption were relaxed and the Dust Bowl temporarily lowered the discount factor, then land
values would decline more in the short-run and increase more in the long-run.
0 is chosen such that 6I'(o) = II'2(1 - 0).
6I (0) + 112(1 -
The farmer earns a long-run profit
LR=
0).
Given an initially binding land allocation constraint, it must be that 0 < 0 and wSR <
L--
rLR < I'.
Profit increases from the short-run to the long-run by f0 IIJ(1 - x) - 6'l(x)d;
intuitively, the difference in marginal profit is regained for each unit adjusted. Taking a
first-order Taylor expansion of each marginal profit function around 0, this term simplifies
to 1( - 0)(1 - 6)('(). The value of this "adjustment triangle" corresponds to one-half
The
the change in land allocation multiplied by the initial decrease in marginal return.
approximation is exact if FI'2(1 - 0) and H1,(0) are linear. As one would expect, profits fall
in the short-run and partially recover in the long-run as the land allocation adjusts.
Change in Land Values. Land values are derived as the net present value of profits,
given a profit stream of 7rSR until period T and rLR thereafter. In each period 0 < t < T - 1,
the value of land is E
rSR3pi
+ E--T 7 LRi; for t > T, the value of land is
LR
This has three implications that will be estimated from the data. First, land values
initially fall by a smaller percentage than profits. This is due to the expected long-run
recovery in profits. Rearranging terms, the value of land at t = 0 is
+
-ir
Z
=TS)
This expression is greater than the PDV of short-run profits (the first term) when the longrun recovery in profits is larger and the long-run is achieved sooner.
Second, land values recover somewhat over time as periods of short-run profits are past
and periods of long-run profits become more immediate. Rearranging terms, the value of
land in each period 0 < t < T -
isR
(rLR
_
SR) E
t 1.
Land values increase in each
period, though by a smaller percent than profits increase when the long-run is reached.
Third, the value of land at t = 0 capitalizes the full PDV economic loss associated with
the shock. Rearranging terms, the value of land at t = 0 is
OT,7rLR+(1_,3T)hSR
L
-l
. Intuitively, this
is a weighted average of long-run profits and short-run profits, where the weights are the
relative value of each.
Change in Land Allocation.
In the long-run, land is allocated away from the pro-
duction technology that experienced a relative decrease in profitability. Rearranging terms
from the profit function Taylor expansion, the change in land allocation (0 - 0) equals the
initial decrease in marginal return ((1 - 3)I1(0)), divided by the summed slopes of the two
technologies' marginal returns at the initial equilibrium (II(1 -
)+
I
"(0)). The change
in land allocation will be greater when there is a larger decrease in marginal return, and
smaller when marginal returns are more sensitive to changes in land allocation.
Change in Revenue and Inputs. There are no general theoretical predictions about
changes in revenue or inputs, because they depend on the functional form of the production
function. However, a Cobb-Douglas functional form provides a useful benchmark: because
factor shares are constant, there would be proportional decreases in profits, revenue, and
inputs.
All major inputs are not reported in the data, so it will not be possible to analyze profits
directly. Rosenzweig and Wolpin (2000) illustrate how unobserved inputs, particularly family
labor, can bias estimates of the effect of environmental shocks on profits (revenues minus
observed input costs). The analysis here will instead focus on percent changes in revenue,
which would proxy for percent changes in profits assuming a Cobb-Douglas functional form.
This assumption is quite strong, so an empirical analysis of observed inputs will explore the
potential direction of bias.
The Effect of Adjustment Costs. Consider now the differences in these effects when
the farmer must pay a non-recoverable cost C 12 (L) to shift land L from technology 1 to
technology 2. For simplicity, assume that the farmer will pay this one-time cost in period T.
Assuming that some land adjustment is still optimal, the new long-run land allocation is
chosen in period T to satisfy the condition: (1 - 6)'1(0) = H'2(1 - 0) - (1 - /)C12(0
-
0").
This condition assumes that the farmer has perfect access to credit so that only (1 - 3) of the
adjustment cost is paid in each period; if instead the farmer faces some capital constraints,
then full initial adjustment will be more costly.
Compared to the model with no adjustment costs, there will be less adjustment in land
allocation. Thus, there will be less long-run recovery in profit, in addition to the direct
reduction in profits from the paid adjustment cost. The value of land will initially decrease
by an additional term, equal to the PDV of the future adjustment cost. This additional term
increases over time as the adjustment cost becomes more immediate. At t = T, the value of
land then increases by the full value of the adjustment cost. 12 The sum total of these effects
is clear in the special case when the adjustment cost is only c less than the PDV long-run
recovery in profits: the value of land decreases at t = 0, remains constant, and then partially
recovers at t = T when it increases by the full amount of the adjustment cost.
If adjustment costs can be fully recovered, such as in purchasing general machinery or
livestock, then the value of land follows the same pattern initially and does not increase at
t = T when these costs are paid. Thus, observed changes in land values can indicate whether
adjustment takes place on margins that require fixed or mobile investments. Note, however,
that this model assumes perfect foresight. Any systematic errors about the future costs of
erosion, potential government subsidies, or other factors will also influence the evolution of
land values.
II.C
(1)
Empirical Predictions for Agricultural Production
The immediate decrease in land values reflects the present discounted value of the
economic cost to agriculture.
(2)
Profits recover partially in the long-run, as fixed production decisions adjust.
(3)
If profits are expected to recover in the long-run, then land values fall by a smaller
percentage than profits in the short-run.
(4)
Under strong functional form assumptions, percent changes in revenue may proxy for
percent changes in profits.
12
If revenue recovers due to an increase in inputs rather than
This refers to the gross value of land, which is reported in the data. The value of land, net of any
mortgage debt incurred, would increase gradually as that debt is repaid.
technological adjustment, then the percent recovery in revenue will overstate the percent
recovery in profits.
Adjustment costs reduce or slow recovery in profits. If adjustment costs are fixed to
(5)
the land, land values increase when the costs are paid.
II.D
A Model of Adjustment Across the Broader Economy
If the Dust Bowl substantially affected agricultural production, this could have general equilibrium effects in non-agricultural sectors and non-Dust Bowl areas. This section outlines
a basic model of supply and demand across locations, sectors, and factor inputs. This is a
model of long-run supply and demand, in which many margins are free to adjust.
In this model, there are two locations (Dust Bowl and non-Dust Bowl). For example, this
could be two counties within Oklahoma that are more- and less-eroded, or one eroded county
in Oklahoma and one non-eroded county in California. There are two sectors (agriculture
and industry) that produce freely tradeable goods using two homogeneous factors (land and
labor). The supply of land is fixed in each location. Labor is supplied by workers who pay a
cost to change location or sector. Workers must supply labor in the same location that they
live. Workers consume land (housing), agricultural goods, and industrial goods. Assuming
perfectly competitive markets, all prices (land values, wages, and prices for agricultural and
industrial goods) are set such that each market clears. 13
For the comparative statics, assume that agricultural productivity declines in the Dust
Bowl area. This decreases the demand for agricultural land. Consider the case in which
the supply of land for agriculture is inelastic, i.e., workers and industrial firms (current or
entering) have little use for additional marginal lands. In this case, adjustment in the land
market mainly occurs through decreased land prices with little change in total farmland.
Assume further that the Dust Bowl decreases the productivity of agricultural labor.
Workers then have an incentive to move to the non-Dust Bowl area and/or switch to the
13This model is similar to that derived by Roback (1982), which focuses on urban amenities.
industrial sector. Equilibrium wages remain relatively lower in the Dust Bowl area (particularly in the agricultural sector) to the extent that there are costs to moving (and switching
sectors). Workers consume some land that is now cheaper, so paid wages fall by more than
local price-adjusted wages.
The industrial sector has an incentive to expand in the Dust Bowl area for two reasons.
First, cheaper land is attractive if the sector has any use for additional lands. Second, the
industrial wage would decrease to the extent that workers switched sectors or consumed
cheaper land. 14 Lower wages encourage higher labor-capital ratios in the agricultural sector,
and discourage adjustment toward production methods that use less labor (such as animals
or hay, relative to crops or wheat).
In the non-Dust Bowl area, wages decrease in response to an in-migration of workers.
Because the Dust Bowl area is less productive, agricultural output prices increase and the
agricultural sector expands in the non-Dust Bowl area. As a result, land prices increase.
The industrial sector contracts due to higher land prices and lower output prices.
Empirical Predictions for the Broader Economy
II.E
(1)
Workers migrate from the Dust Bowl area, though this incentive is tempered by lower
housing prices and moving costs. If workers migrate to the compared non-Dust Bowl area,
the relative comparison in population will overstate the number of migrants.
(2)
A greater share of workers become employed in the industrial sector, which also ex-
pands due to lower land values and wages.
(3)
If industry and workers have inelastic demand for agricultural land, then adjustment in
land markets occurs mostly through lower prices rather than a contraction in total farmland.
(4)
Lower wages encourage labor-intensive production in the agricultural sector.
14The industrial sector could contract if it was supplied inputs by the local agricultural sector, or if it
sold outputs locally to the agricultural sector. This is not captured in the model, but could be empirically
relevant.
(5)
Decreased agricultural production in the Dust Bowl area raises land values in the
non-Dust Bowl area, causing the relative comparison of land values to overstate the total
cost of the Dust Bowl.
(6)
Discouraged production activities in the Dust Bowl area are encouraged in the non-
Dust Bowl area, due to increased output prices. Relative comparisons in production across
sectors or agricultural products would overstate adjustment in the Dust Bowl area.
III
III.A
Data and Summary Statistics
Measuring Erosion
This paper uses new detailed data on which areas were most eroded after the 1930's, which
I gathered by digitizing soil erosion survey maps currently in the National Archives. In
response to the Dust Bowl, these maps were produced by the Soil Erosion Service (SES)
and its successor, the Soil Conservation Service (SCS). The SES published the 1934 Reconnaissance Erosion Survey, which detailed the severity and type of erosion across the United
States based on measurements taken during visits to each county (SCS 1935). In August
1936, the SCS published a map of which general areas had been affected to different degrees
and a second more-detailed map in December 1937.
This first type of erosion map indicates whether areas were slightly, moderately, or
severely eroded; whether it was wind erosion (topsoil being swept from the land) or sheet
erosion (water generally shifting the topsoil off the land); and whether there were many
gullies (formed by small streams of runoff carving paths in the land). There are two main
limitations to this map type. First, it only reflects the stock of erosion at the time of the
survey. Since there is no detailed baseline data and the erosion category definitions change
over time, it is not possible to measure directly the erosion that occurred during the Dust
Bowl. Second, it is difficult to compare erosion across categories and generate a single index
for how much an area was affected.15
The SCS prepared a second type of map that indicates dated regions of wind erosion
on the Plains. These show broad areas that experienced blowing from 1935 to 1936 and in
1938, which correspond closely to areas with the highest wind speeds (Chepil et al. 1962).
These regions have less local variation, however, and it is not clear how the exact areas were
determined. There was substantial wind erosion at other times during the Dust Bowl, as
well as water erosion, and these maps give little sense of the cumulative effect on the soil.
My analysis focuses on a third type of SCS erosion map, which shows cumulative erosion
damage after the Dust Bowl. The map classifies areas as slightly eroded (less than 25% of
topsoil lost), moderately eroded (25% to 75% of topsoil lost), and severely eroded (more
than 75% of topsoil lost). Of all the SCS erosions maps, this map has the most detailed
local variation in erosion. Documentation is sparse, but this map (along with most others)
is recorded as "based on the 1934 Reconnaissance Erosion Survey and other surveys." The
National Archives have three copies of this same map with publishing dates in 1948, 1951,
and 1954 - prior to the next substantial period of erosion in the mid-1950's.
The main limitation of this cumulative erosion map is that it does not just reflect erosion
that occurred during the 1930's Dust Bowl. Total erosion would have begun to increase
along with the settlement of the Plains. Large-scale detailed erosion surveys only began in
the 1930's, so there is no systematic baseline data on erosion and topsoil levels prior to the
Dust Bowl. To focus on the change in erosion over the 1930's, the empirical analysis controls
for a range of county characteristics in 1930, 1925, 1920, and 1910 that might predict baseline
erosion levels.
To focus on the extreme erosion of the 1930's, the sample is restricted to counties within
the 12 Plains states: Colorado, Iowa, Kansas, Minnesota, Montana, Nebraska, New Mexico,
15
Hansen and Libecap (2004) assigned erosion categories based on the 1936 map, as they are focused on
specifically wind erosion.
North Dakota, Oklahoma, South Dakota, Texas, and Wyoming. 16 In the cumulative erosion
map, many areas in non-Plains states were also classified as moderately or severely eroded.
Non-Plains states had only severe water erosion indicated in other maps, however, which is
more likely to reflect gradual erosion over many previous decades. As a robustness check,
the empirical analysis shows that more-eroded non-Plains counties did not experience the
substantial relative drop in land values during the 1930's that occurred in more-eroded
Plains counties. This suggests that the cumulative erosion measure is an effective proxy for
the extreme Dust Bowl erosion on the Plains in the 1930's.
III.B
Combining County Outcomes with Erosion Data
The core dataset is compiled from US Census data on agriculture, population, and industry,
aggregated to the county level. Among the key variables are the value and quantity of
agricultural land, agricultural revenue, total cropland and pasture, revenue from crops or
animals, production and acreage for a variety of crops, the number of farms, population,
rural population, farm population, retail sales, manufacturing workers and value added,
and unemployment."
Other data sources include banking data from the FDIC; New Deal
expenditures from the Office of Government Reports; and drought data from the National
Climatic Data Center.18
The data are combined into a panel of US counties, where the unit of observation is
held fixed at 1910 county boundaries.
16
Census data are used from 1910, 1920, and then
The results are similar when the sample is restricted further to "Great Plains counties," as defined by
the Great Plains Population and Environment Database (which provided additional historical data on these
12 states). This reduces the sample from 769 counties to 356 counties concentrated in western Kansas and
Oklahoma, northwest Texas, central and western Nebraska, eastern Colorado, and the Dakotas.
17
An unusual aspect of this data concerns the calculation of agricultural revenue in 1920, 1925, and 1930.
In these periods, revenues from animals sold and animal products sold are not reported. Only the total value
of animals is reported. In 1910 and 1940, both of these variables are reported. The ratio of revenue to stock
is calculated for each county and used to impute the county's value of animal revenue in the intervening
years. The results are not sensitive to using the revenue/stock ratio in 1910, 1940, or a weighted average of
the two - the results presented here use a weighted average (with the 1910 ratio weighted by two-thirds in
1920, one-half in 1925, and one-third in 1930).
18I thank Price Fishback for providing the New Deal data (Fishback et al. 2005) and the drought data
(Boustan et al. 2007).
approximately every 5 years through 1997, though data availability for some variables is
more restricted. Data from 1935 is dropped, due to ongoing erosion from the Dust Bowl.
When counties split or borders were otherwise adjusted in subsequent periods, the reported
Census data are aggregated or adjusted to reflect the original 1910 county border (Hornbeck
2008). The base year was chosen as 1910 in a tradeoff between more sample counties and
a longer pre-period for analysis: changing from 1900 to 1910 substantially increases the
number of sample counties in Oklahoma and other areas, while moving beyond 1910 has less
effect on sample size and the pre-period becomes less informative about differential trends.
The core results focus on the same sample of counties, for which data on key variables are
available in every period of analysis. 19 This allows for a clearer interpretation of estimated
changes over time and across specifications, because the sample composition is not also
changing.
Figure 2 displays these 769 sample counties, along with the mapped cumulative erosion
levels. The white area represents low erosion, the light gray is medium erosion, and the
dark gray is high erosion (the crossed out areas are not in the sample). 2 0 The figure was
constructed by overlaying the cumulative erosion map on a map of United States historical
county borders using Arcview Geographical Information System (GIS) software. To create
the data, the erosion areas were traced and merged with 1910 county borders to assign each
county the percent of its area that falls within each erosion category.
19In a tradeoff between timespan and sample size, the sample is restricted to counties that have data on:
land values, revenue, and farmland through 1992; cropland through 1974; population and rural population
through 1969. The sample also excludes a small number of counties with less than 1000 acres of farmland
and a few counties with extreme values due clearly to measurement error and/or the adjustment to maintain
constant boundaries.
20
The main reasons why areas are dropped from the sample are: the area was not sufficiently settled by
1910 for data to be reported; data are unavailable in 1925; agricultural revenue could not be imputed for
1920, 1925, or 1930; and border adjustments left data unavailable for any piece that made up more than 1%
of the original county's total area.
III.C
Summary Statistics, by Erosion Level
Table 1 presents baseline differences between sample counties in 1930, based on their assigned
post-Dust Bowl erosion level. 21 For comparison, column 1 presents mean characteristics for
all counties in 1930. Columns 2 and 3 report coefficients from a single regression of each
outcome variable on the percent of a county in medium erosion and high erosion, controlling
for state fixed effects. 22
Counties with more erosion after the Dust Bowl tended to have previously: higher land
values, denser population, more but smaller farms, a larger fraction of cropland in corn or
cotton (as opposed to wheat and hay), and more animals. Column 4 reports the average
difference between a county with high erosion and a county with medium erosion. These
counties were more similar, though high erosion counties have more cropland allocated to
corn.
Measured erosion levels are not randomly assigned for two reasons. First, part of this
erosion occurred prior to the Dust Bowl and could be caused by, reflected in, or otherwise
jointly determined with pre-Dust Bowl county characteristics.
Second, the Dust Bowl's
intensity was partly determined by the intensity of cultivation and other county land-use
practices. The main empirical challenge is that counties with these different characteristics
may have changed differently after the 1930's, even if the Dust Bowl had not occurred.
To address this issue, the empirical analysis will focus on specifications that control for
differential changes over time that are correlated with state, county characteristics in 1930,
and lagged values of those characteristics. The empirical analysis will not control for county
characteristics after 1930, as these are potential outcomes of the Dust Bowl. A robustness
check will instrument for county erosion levels using intensity of drought during the 1930's.
21
Given the construction of the sample, there are no missing values for panels A, B, and C. For panels D
and E, missing values are assumed to be zero (generally in areas where these products are not produced).
22
The means and regressions are weighted by county farmland levels, as the empirical analysis will be
focused on the effect for an average acre of farmland. These variables will all be included as controls in later
weighted regressions, so their weighted difference is more relevant to interpreting their importance as control
variables.
IV
Empirical Framework
The empirical analysis is based upon comparing changes in outcomes for counties that experience different levels of erosion. For a graphical analysis of the data over the entire sample
period, outcome Yt in county c and year t is regressed on the percent of the county in
medium erosion (Me) and high erosion (He) areas, a state-by-year fixed effect (at), linear
functions of county characteristics (Xc), and an error term (eat):
(1)
YMt =
ltMc +
ost + OtXc + Ect.
+
±2tHe
The effects of erosion and each county characteristic are allowed to vary in each year. The
included county characteristics are each of the variables in panels B - E of table 1, and their
lagged value from all available previous years. In each regression, every county included has
data in every analyzed period.
This regression is a special case, in which each regressor is fully interacted with each time
period. Thus, controlling for a county fixed effect would not change the difference between
estimated P's; rather, it simply normalizes the p's relative to zero in some base period.
County fixed effects are omitted in this regression in order to observe the pre-Dust Bowl
difference in outcome levels across counties with different erosion levels, after controlling for
the other pre-Dust Bowl characteristics.
Graphing the estimated p's shows how counties with each erosion level changed over
the sample period, relative to a county with low erosion. To interpret these changes as the
relative effect of the Dust Bowl, the primary identification assumption is that counties with
different erosion levels would have changed the same if not for the Dust Bowl. In practice,
this assumption must hold after controlling for differential changes over each time period
that are correlated with each state and pre-Dust Bowl county characteristic.
From the estimation of equation (1), it is possible to observe whether counties with
different erosion levels were on similar trends from 1910 to 1930 (after controlling for other
covariates). Because counties may not be on the same trend prior to the Dust Bowl, the
empirical analysis then controls for these differences in pre-trends.
For estimating the numerical results, two modifications are made to equation (1). First,
the regression controls for the outcome level in all available pre-Dust Bowl periods, L,. This
is a flexible way to control for differential pre-trends, without assuming that trends prior to
the 1930's would have continued predictably - which is theoretically inappropriate for asset
prices such as land values, and is empirically questionable for other outcomes.
Second, the outcome Yt is differenced from its value in 1930, so that each coefficient can
be interpreted as the relative change since 1930. Differencing the data also improves the
precision of the estimates by absorbing any fixed county characteristics; yet, as before, the
estimated coefficients are not changed by differencing the data or by including a county fixed
effect.23 The regression is otherwise the same:
(2)
Yct
-
Yc1930 -=
1tM
+ /2tHe
+ st + OtXc + 7tLc
+ Ect.
As before, note that the effects of erosion and each county characteristic are allowed to vary
in each year. In order to condense the results, a specification is sometimes estimated that
pools the erosion variables for some combination of later time periods. The control variables
are still allowed to vary in each year, so this amounts to averaging the estimated '3's across
years.
Regressions for agricultural outcomes are weighted by county farmland (or an analogous
land measure) in 1930 in order to estimate the average effect for an acre of farmland. Regressions for labor outcomes are weighted by county population in 1930 in order to estimate
the average effect for a person. Standard errors are clustered at the county level, to adjust
for heteroskedasticity and within-county correlation over time. To check for spatial correla23
Because the sample is balanced and the regressors are fully interacted with each time period, estimation
of the regression is completely separable across year pairs. The methods are equivalent in the case of two
time periods. While the coefficients are identical, the methods have slightly different standard errors. This
differencing was done throughout for ease of interpretation and computational speed.
tion among counties, Conley standard errors are estimated for changes in land values from
1930 to 1940 (Conley 1999). These standard errors are similar to standard errors clustered
at the county level, indicating that the impact of the Dust Bowl was not highly spatially
correlated. 2 4
V
Results
V.A
Agricultural Land Value and Revenue
Agricultural Land Value. Figure 3 graphs the 3's from estimating equation (1) for the
per-acre value of farmland. High and medium erosion counties had slightly higher land values
than low erosion counties in 1910, after adjusting for the control variables. These counties
were more similar by 1920 and then trended similarly through 1930.
After the Dust Bowl, high erosion counties experienced an immediate, substantial, and
persistent relative decrease in land values. Medium erosion counties experienced a smaller but
substantial relative decrease. Comparing the two lines, high erosion counties also experienced
a persistent decrease relative to medium erosion counties.
To account for differences in the level and trend of county land values prior to the 1930's,
the numerical results are obtained by estimating equation (2). This regression controls for
county land values in 1930, 1925, 1920, and 1910, interacted with each year.2 5 These controls
absorb the graphed pre-1930's differences by erosion level, and are a flexible way of adjusting
for differential pre-trends in land values. Table 2, column 1, presents the results.
24
The Conley method allows for outcomes to be correlated among nearby counties, with the degree of
correlation declining linearly until some cutoff distance. Relative to when there is restricted to be no crosscounty correlation, the percent increase in standard errors for changes in high erosion vs. low erosion counties
is: 0% (50 mile cutoff), 5% (100 mile), 16% (300 mile), 30% (500 mile), 29% (700 mile), 23% (900 mile). For
changes in medium erosion vs. low erosion counties, the Conley standard errors decline by 0% to 17%. The
Conley specifications are not also weighted by county farmland levels, so the comparison in standard errors
is relative to when not weighting the regression.
25
In this case, the regression also controls for agricultural revenues per-acre in each pre-period, as the later
analysis will combine the specifications in a seemingly unrelated regression framework.
Panel A reports that, from 1930 to 1940, land values fell by 27.8% in high erosion counties
relative to low erosion counties. Land values fell by 16.7% in medium erosion counties relative
to low erosion counties. If we assume that low erosion counties were not affected by the Dust
Bowl, the total capitalized cost can be approximated by multiplying the percent decline in
land values by the original value of those acres. These estimates indicate that the Dust
Bowl erosion imposed an economic cost on agriculture of $153 million in 1930 dollars ($1.9
billion in 2007 dollars).2 6 If low erosion counties were also damaged by the Dust Bowl, this
number would understate the total cost. If the Dust Bowl made less-affected land in the
same state more valuable, e.g., through higher output prices, this number would overstate
the total cost. 27 Given average 1930 land values, this reflects an area of farmland the size of
Oklahoma becoming worthless.
Panel B reports the estimated change in land values from 1930 to 1945.
Column 2
expresses this change as a fraction of the change in land values from 1930 to 1940. These
numbers reflect the amount of decrease from 1930 to 1940 that persisted into 1945: 84% for
high erosion and 61% for medium erosion. For changes in high erosion counties relative to
medium erosion counties, the persistence was 119% (land values declined further). There
is no strong a priori reason for one of these three relative comparisons to reflect a more or
less persistent shock. Taking the efficient weighted average of these three parameters, the
average persistence is 76.5%.2"
26
By multiplying 1930 county farmland levels by the share of its area in each erosion category, there were
approximately 51 million farm acres in high erosion areas and 170 million farm acres in medium erosion areas.
The per-acre value of farmland was $3.60 in high erosion counties and $3.62 in medium erosion counties,
weighting by farmland. The total cost is found by multiplying the acres affected, the value of the acres, and
the percent decline in land values: $51 million from high erosion and $102 million from medium erosion.
27
There is no obvious indication of higher prices from national trends in the ppi for farm products and
the cpi for urban consumers: farm output prices were lower from 1938 to 1940 than in the 1920's - both
absolutely and relative to the urban cpi. All prices declined in the 1930's and increased during WW2, while
farm output prices increased during the severe droughts from 1934 to 1936.
28
The variance-covariance matrix for these three estimates is known, so the efficient average is estimated
through GLS (regressing the three values on a constant). Intuitively, this procedure gives more weight to
the more-precisely estimated coefficients, and less weight to those coefficients are are more correlated with
each other.
Panels C, D, and E report later changes in land values, pooling the estimates over 2, 3,
and 4 census periods. There is little indication of a systematic recovery in land values. This
suggests that the overall percent recovery in profits is not large, and that adjustment did
not take place through fixed improvements in land that would be capitalized in its value.
Agricultural Revenue. Figure 4, panel A, graphs the O's from estimating equation (1)
for the per-acre value of farm revenue. From 1930 to 1940, more-eroded counties experienced
substantial declines in farm revenue. Revenue recovered partly from 1940 to 1945 and then
the paths diverge: high erosion counties have lower revenue compared to low erosion counties,
while revenue gradually recovers in medium erosion counties. Revenue had been increasing
in more-eroded counties from 1910 to 1925, and then decreased from 1925 to 1930. Table
2, column 3, presents the numerical results for farm revenue, controlling for these pre-1930
differences.
Panel A reports the relative decrease in farm revenue from 1930 to 1940: 31.6% for high
erosion counties and 20.2% for medium erosion counties. Panels B to E report the changes
for later periods, and column 4 expresses these changes as a fraction of the decrease from
1930 to 1940. From 1940 to 1945, more-eroded counties experienced a substantial recovery in
revenues, though averaged levels were still lower than in 1930. Over later periods, however,
only medium erosion counties experienced a substantial recovery in revenues. On average,
approximately 70% of the initial decline in revenues persisted between 1950 and the 1990's.
One way of interpreting these results is to take the present discounted value of all lost
revenue after 1940. This calculation assumes an interest rate of 5%, linearly interpolates
annual values between each census period, and assumes that revenue is constant after 1992.
To obtain a measure of cost persistence, this PDV is expressed as a fraction of the PDV
of lost revenue assuming no recovery after 1940.29 The estimated persistence in losses (and
standard error) is: 0.762 (0.132) for high erosion relative to low erosion, 0.687 (0.152) for
medium erosion relative to low erosion, and 0.896 (0.351) for high erosion relative to medium
29
To do this calculation, annual coefficients for column 4 are linearly interpolated. Given an interest rate
r, the estimated persistence in year t is multiplied by
These valus are summed and divided by
erosion. If percent changes in revenue were a proxy for percent changes in profits in every
period, then these numbers represent the full economic loss as a fraction of the short-run
loss. However, if the recovery in revenue is due to an increase in production inputs, then
these numbers will understate the persistence in lost profit.
The value of capital machinery and equipment is the primary production input for which
comparable data are available across periods in all sample counties. Figure 4, panel B, graphs
the results from estimating equation (1) for the log value of capital per acre of farmland.
Capital inputs fell substantially from 1930 to 1940, though not by as much as revenues. 30
Capital inputs partially recovered from 1940 to 1945 and then diverged: inputs increased
over time for medium erosion counties and declined again for high erosion counties. Thus,
part of the observed recovery in revenue may be due to greater inputs, overstating the
recovery in profits - particularly for the medium erosion counties in which revenue recovered
substantially.
Agricultural Land Value vs. Revenue.
Comparing the relative changes in land
values and revenues provides a second estimate of the persistence in short-run costs. The
percent decline in land values should reflect the true economic loss from the Dust Bowl,
anticipating and appropriately discounting the future recovery in profits.
Under strong
functional form assumptions, the immediate percent decline in revenue would proxy for
the immediate percent decline in profits. The true economic loss (land value) can then be
expressed as a fraction of the short-run economic loss (revenue or profit).
Column 5 of table 2 reports this ratio, estimated for each relative comparison between
erosion levels.3 1 If the theoretical assumptions hold, these ratios have a clear structural interpretation: for every 1% lost in the short-run (1940), the full economic loss is between 0.82%
and 0.98%, with a weighted average of 0.86%. Theoretically, this ratio should be weakly
less than one, though the estimated parameters need not be. Because none are statistically
30
From estimating equation (2) for capital inputs, the decrease was 0.225 (0.055) for high erosion and
0.121 (0.037) for medium erosion, relative to low erosion.
31
To obtain the standard error of this ratio, both coefficients are estimated through seemingly unrelated
regression, where the same control variables are included in each regression.
less than one, these estimates fail to reject the hypothesis that long-run adjustments did
not mitigate the short-run costs. These coefficients are all statistically greater than zero,
which rejects the hypothesis that short-run costs were associated with no PDV cost. If the
theoretical assumptions fail to hold, these numbers still have an appealing reduced-form
interpretation: the full economic loss (land value) is scaled by the magnitude of the shock
(revenue).
The main limitations in interpreting this ratio are the strong assumptions required for
immediate percent changes in revenue to proxy for immediate percent changes in profits.
Capital inputs fell substantially from 1930 to 1940 but somewhat less than revenues, so
profits may have fallen by a larger percent than revenues. In this case, the estimated ratios
in column 5 would overstate the persistence in costs (the denominator should be larger). By
contrast, the persistence in costs may have been understated by the estimated recovery in
revenues because capital inputs increased after 1940. The difference between revenues and
profits may then bias the two cost persistence estimates in opposite directions. The range
of cost persistence estimates from the two approaches is: between 0.762 and 0.881 for high
erosion vs. low erosion, between 0.687 and 0.824 for medium erosion vs. low erosion, and
between 0.896 and 0.981 for high erosion vs. medium erosion.
V.B
Measured Erosion as a Proxy for the Dust Bowl
The empirical analysis relies on mapped erosion levels indicating areas that were more or
less eroded during the Dust Bowl, after adjusting for state fixed effects and pre-Dust Bowl
county characteristics. This section presents further results on changes in land values, in
order to explore the validity of these erosion measures.
Plains vs. Non-Plains. One concern is that some of the measured variation in erosion
levels occurred prior to the 1930's. Indeed, the nationwide map of cumulative erosion indicates that there were many severely eroded areas in the Eastern United States, where the
Dust Bowl did not cause additional erosion. If counties with different baseline erosion levels
would have changed differently over the 1930's, this would be confounded with the impact
of the Dust Bowl.
To explore this possibility, figure 5 compares estimated relative changes in land values
for more-eroded counties in Plains and non-Plains states. For each year, panel A (panel
B) graphs the estimated difference in land values between high and low erosion counties
32
(medium and low erosion counties), controlling only for state-by-time fixed effects.
Before 1930, more-eroded Plains counties have higher land values than less-eroded Plains
counties; and this difference is greater than the difference between more-eroded and lesseroded counties in non-Plains states. One explanation is that "more-eroded" Plains counties
had yet to experience much of that erosion prior to the Dust Bowl. Following the Dust
Bowl, more-eroded counties on the Plains experience a large relative decrease in land values.
By contrast, land values in more-eroded and less-eroded counties in non-Plains states were
relatively unchanged. These counties are otherwise on similar relative trends before 1930
and after 1960, at which point the level differences are also similar.
These estimates have two implications. First, the cumulative erosion measure appears to
proxy for the particularly destructive erosion on the Plains during the Dust Bowl. Second,
differentially eroded counties experienced similar changes in land values over the 1930's when
the Dust Bowl did not contribute to further erosion.
Instrumenting for Erosion with Drought. A related concern is that observed erosion
levels partly reflect counties' land-use, both prior to 1930 and during the Dust Bowl. When
analyzing the effect on land value, OLS estimates could be negatively biased if farmers in
counties otherwise becoming less valuable did not protect their land from Dust Bowl erosion.
OLS estimates could be positively biased if counties otherwise becoming more valuable were
farmed more intensely and this caused greater Dust Bowl erosion. Measurement error in
erosion levels could also attenuate the estimated effects.
32
The other control variables are not available for all non-Plains counties. The sample is expanded to
include all Plains and non-Plains counties with available land value data in each period: 867 Plains counties
and 1840 non-Plains counties.
Some portion of the Dust Bowl erosion was due to severe drought on the Plains during the
1930's. As a robustness check, the analysis instruments for erosion levels using the number
of months during the 1930's that a county was in extreme drought, severe drought, moderate
drought, and the average level of drought over the 1930's.33 The analysis now also controls
for these four drought variables from 1895 to 1929, in order to capture general differences in
susceptibility to drought.
The identifying assumption is that 1930's weather was not systematically correlated with
future changes in land values, aside from its impact through changing erosion. These drought
variables are sometimes correlated with pre-1930's county characteristics, so the empirical
analysis continues to control for these variables. Temporary weather shocks should not have
a direct effect on land values, though fixed investments in land could decrease in response
to drought if income falls and credit is difficult to obtain. 34
Table 3, columns 1 and 2, report the first-stage results. Column 1 reports estimates
from regressing the percent of a county in high erosion areas on the drought instruments,
controlling for drought from 1895 to 1929, state fixed effects, and the same pre-1930's county
characteristics as before. There is more high erosion in counties with more months of extreme
drought, and a weaker effect in counties with more severe drought.
High erosion is not
predicted by months of moderate drought or average drought over the 1930's. Column 2
reports analogous estimates for the percent of a county in medium erosion areas. Medium
erosion is higher in counties with higher average drought levels and more moderate drought,
with slightly weaker effects from severe and extreme drought. The instruments are jointly
significant, but the overall explanatory power is low and raises weak instrument problems.
35
Ongoing work is gathering additional potential instruments.
33
Drought is measured by the Palmer Drought Index, which is defined on a running basis depending on
the arrival of rainfall and the loss of moisture. Extreme, severe, and moderate reflect conventional cutoffs
in this Index. The drought data were collected at weather stations, and the PDI is reported by state and
district (10 districts to a state). Details are available at ftp://ftp.ncdc.noaa.gov/pub/data/cirs/
34
This exclusion restriction appears more likely to be violated for other county outcomes, so the instrumental variables specifications are only estimated for changes in land values.
35
Average wind speeds may predict wind erosion, and slopes may predict water erosion. There are direct
measures of soil characteristics and its susceptibility to erosion, but these data are only available in the
Table 3, column 3, reports the estimated relative changes in land values from 1930 to
1940, instrumenting for erosion levels with the 1930's drought variables.
As a basis for
comparison, column 4 presents the OLS results for the same sample and control variables.
When instrumenting, higher erosion counties experienced a larger relative decline in land
values (63% instead of 28%) and medium erosion counties experienced a somewhat smaller
relative decline (12% vs. 17%).
The implied cost of the Dust Bowl increases from $153
million to $187 million (in 1930 dollars). These IV estimates are much less precise, but they
are consistent with the previous OLS findings: measured erosion levels appear to proxy for
the impact of the Dust Bowl and decreased land values. 36
V.C
Adjustment in Agricultural Production
Agricultural costs from the Dust Bowl were estimated to be persistent. There are three main
interpretations for this finding. The first interpretation, which later estimates support, is
that adjustment in agricultural production was possible but costly and slow. The second
interpretation is that adjustment to the Dust Bowl was not possible. The third interpretation
is that adjustment was fast and had already taken place by 1940.
This section attempts to identify examples of possible margins of adjustment, and then
what adjustments occurred. Results from estimating equation (1) are graphed in Figures 6
- 9, in order to observe relative pre-trends in each agricultural outcome. Table 4 reports
the numerical results from estimating equation (2), which controls for differences in relative
pre-trends.
Total Farmland. Figure 6 graphs the estimated changes in total farmland. This extensive margin of farming is fairly stable: immediately after the Dust Bowl, there are no submodern era and may be an outcome of the Dust Bowl erosion. Hansen and Libecap (2004) use these average
wind speeds and soil characteristics as instruments. The instruments could potentially be interacted with
each other or pre-1930 county land-use. Because the instruments must predict the transition from low erosion
to medium and then high erosion, non-linear functions of the instruments may be appropriate. In the case
of the drought instruments, including their squared values yields similar 2SLS results.
36
As a falsification test, the same specification can be used to examine pre-differences in land values:
more-eroded counties have similar or lower land values than in 1930, instrumenting for erosion with 1930's
drought. In periods after 1940, the estimates are quite variable but fluctuate around their value in 1940.
stantial or statistically significant relative changes in farmland. These small relative changes
in total farmland suggest that the previous estimates of changing land values, revenue, and
capital inputs do not reflect large changes in the underlying composition of farmland.
High erosion counties later experience a gradual small decline in farming, which by the
1950's is lower by a statistically significant 3% (table 4, column 1). These declines in farming
are then largely reversed, however, and medium erosion counties are relatively unchanged. 37
One explanation for the persistence in overall land allocation is that non-agricultural
sectors of the local economy might have an inelastic demand for land, at least in the shortrun. In this case, the supply of land to the agricultural sector would be relatively inelastic.
When agricultural demand for land decreased following the Dust Bowl, we would then expect
large declines in land values and little change in the extensive margin of farming.
After the original settlement of the Plains slowed, the aggregate allocation of land to
agriculture was fairly constant (figure 1). Assume that aggregate demand for agricultural
land is constant, and that county-level fluctuations are driven by county-specific demand
shocks for non-agricultural land. If these demand shocks have little effect on the county
allocation of land to agriculture, then the agricultural sector may face a relatively inelastic
supply of land. Indeed, the standard deviation of county-level fluctuations is relatively small
(0.057).3 8
There may be something about these counties in the 1930's and 1940's that
limited adjustment, and later analysis will explore the possibility of credit constraints and
other factors.
Crops vs. Animals. Higher quality land is generally thought to have a comparative
advantage for growing crops, as opposed to raising animals. There is a reasonable a priori
expectation that erosion from the Dust Bowl would reduce the productivity of land for crops
by more than for animals. Thus, we might expect long-run conversion of land from growing
37
These estimates do not reflect an increase in abandoned farmland, proxied by the difference between
total farmland and cropland or land in pasture. Relative changes in this measure are never greater than 1%
of the total land in farms, in either direction.
38
This is calculated by first-differencing county farmland shares from 1945 to 1992, and taking the standard
deviation of these differences.
crops to raising animals. Indeed, these arguments were made strongly by contemporaries,
and this land conversion was a major goal of government policy, advocated as both good for
the farmer and good for society.3 9
To explore this potential margin of adjustment, it is possible to estimate relative changes
in the productivity of land for crops and animals. This has three main limitations. First,
changes in relative productivity need not imply changes in relative profitability if input
quantities or prices also change. Second, it is only possible to estimate changes in the productivity of the average unit of land, whereas adjustment incentives depend on the marginal
unit of land. Third, if land allocations adjust, then changes in land composition would be
confounded with productivity changes.
Crop productivity is defined as the total value of crops sold, divided by acres of cropland.
Animal productivity is defined as the total value of animals sold and animal products sold,
divided by acres of pasture. These measures of productivity are problematic because some
unknown portion of cropland would be used to grow crops that are fed to that farm's animals.
The analysis would overstate the relative decline in crop productivity if farmers in higher
erosion counties became more prone to use cropland to feed their own animals.
Figure 7, panels A and B, present the relative changes in crop and animal productivity for
high erosion and medium erosion counties, respectively. 40 Both productivities are normalized
to zero in 1930, in order to allow a clearer comparison between the relative changes. For
both high erosion and medium erosion counties, crop productivity declined more than animal
productivity from 1930 to 1940. This relative decrease only persists for medium erosion
counties. 4
39
Table 4, columns 2 and 3, report the corresponding numerical results. Crop
The SCS in 1955 was still strongly advocating conversion of land from cropland to grassland (Allred and
Nixon 1955).
40
The crop and animal productivity regressions are weighted instead by 1930 levels of county cropland
and pasture, respectively. This allows the estimates to be interpreted as the percent change for an average
unit of that land.
41
Note that there was no relative change in productivity for high erosion counties relative to medium
erosion counties. This may reflect a particular functional form through which erosion affects crops. This
is consistent with earlier results on land values and revenue, where the high vs. medium shock had more
persistent costs - there may be fewer opportunities to adjust production.
productivity appears to have been more sensitive to erosion than animal productivity. The
exception is for high erosion counties after 1945, in which animal productivity also declined
substantially.
However, farmers did not begin systematically shifting land from cropland to pasture until
the 1950's (figure 7 panel C and D, table 4 column 4). High and medium erosion counties
both then shifted land. These estimates are consistent with adjustment costs declining in
the long-run.
Counterintuitively, the largest shift occurred in high erosion counties, for
which there were not estimated relative productivity differences in later periods. This may
reflect poor quality land being shifted to pasture, which would account for the accompanying
decrease in pasture productivity. Alternatively, the quality of the empirical comparison may
be deteriorating over time.
Wheat vs. Hay. A similar empirical exercise compares relative changes in the productivity of wheat and hay, and the relative allocation of land to each. These are the two
most widely grown crops for which comparable data are available over a long period of time.
There is also an a priori expectation that erosion would discourage the production of wheat,
relative to hay. Wheat production is more sensitive to soil quality and more likely to cause
erosion, whereas hay is cultivated grass. Hay is less sensitive to soil quality, less prone to
cause erosion, and an input in the production of animals. In contemporaneous and later
writings, it was generally recommended that farmers shift land from wheat to hay (or to
native grasslands and pasture).
Productivity for each crop is defined as the total quantity produced divided by the total
acreage harvested.
Because acreage is only available for harvested land, complete crop
failure would cause the analysis to understate declines in productivity. Many counties do
not report wheat data for 1940, but the reason is unclear.4 2 The same caveats as before apply:
productivity may not reflect changes in profitability; changes in the average acre may not
proxy for changes in the marginal acre; and compositional changes may affect productivity.
42
Data are often unavailable in one of the analyzed periods, so restricting the analysis to a balanced panel
reduces the sample size by roughly half.
Figure 8, panels A and B, present the relative changes in wheat and hay productivity
for high erosion and medium erosion counties, respectively.43
From 1930 to 1940, wheat
productivity decreased substantially. In 1950, when both production and acreage data are
again surveyed for hay, wheat productivity remains persistently less than hay productivity.44
Columns 5 and 6 of table 4 report the numerical results.
Wheat productivity recovers
substantially after 1964, but it remains relatively lower than hay productivity.
Figure 8, panels Panels C and D, show that farmers did not begin systematically shifting
land from wheat to hay until the 1950's or 1960's. Table 4, column 7, reports the numerical
results. After 1964, there were substantial and statistically significant declines in the amount
of land devoted to wheat, as a fraction of the total land devoted to wheat and hay. Much
of this adjustment occurs after data is no longer available for cropland and pasture, so it is
difficult to compare the amount of adjustment along each land-use margin.
Average Farm Size and Land Fallowing. Hansen and Libecap (2004) present evidence that the presence of small farms limited efforts to prevent wind erosion during the
Dust Bowl, due to substantial externalities. If more-eroded counties remained more prone to
wind erosion, one potential margin of adjustment would be the consolidation of land holdings to internalize these externalities. Indeed, estimates indicate that more-eroded counties
experienced greater increases in average farm size during the 1940's (figure 9 panel A, table
4 column 8). However, there may be other reasons for farm sizes to adjust.
If increased farm sizes primarily reflected a desire to incorporate these externalities,
there should be a corresponding increase in efforts to prevent erosion. Previous estimates
show that cropland did not decrease relatively in more-eroded counties during the 1940's.
As another method of preventing erosion, Hansen and Libecap (2004) analyze the fraction
of cropland kept fallow. More-eroded counties, however, experienced relative decreases in
43
The wheat and hay productivity regressions are weighted by 1930 acres of land devoted to that crop.
The other crop and animal variables are omitted as controls, as they directly change along with wheat and
hay.
44
As in the case for crop and animal productivity, there is not much relative change between high and
medium erosion counties.
fallowing during the 1940's (figure 9 panel A, table 4 column 9).
It does not seem that
more-eroded counties relatively adjusted production to prevent future erosion.
These estimates imply that more-eroded counties became more intensely cultivated from
1940 to 1945, similar to the estimated increases in capital inputs. This would cause the
observed percent recovery in agricultural revenues to overstate the percent recovery in profits.
V.D
Why Was Adjustment in Agricultural Production Limited?
This section explores three possible explanations for the observed slow and limited adjustment in agricultural production: credit constraints, low incentives for tenants to adjust, and
government payment programs that might discourage adjustment.
Credit Constraints. During the Great Depression, access to credit may have been
generally restricted. In particular, counties with higher erosion experienced substantial decreases in land values, so land-owning farmers would have lost potential collateral. Poorly
performing local mortgages may have restricted banks' ability to lend. The most-eroded
counties also experienced more bank failures during the 1930's (figure 10). Bank weakness
and bank failures can lead to persistent decreases in the supply of credit, especially during
the 1930's (Bernanke 1983, Calomiris and Mason 2003, Ashcraft 2005).
Without easy access to credit, it may have been difficult for farmers to adjust agricultural
production in response to the Dust Bowl erosion. Raising animals or shifting crops would
require the purchase of livestock and perhaps different machinery. Other local sectors would
require capital to shift land from agriculture to industrial uses or residential housing.
To explore whether restricted access to credit might explain the limited adjustment in
agricultural land-use, I estimate whether counties affected by the Dust Bowl adjusted more
if they had more banks prior to the 1930's. 45 The estimated equation is a modified version
of equation (2). It adds interaction terms between the log number of banks (B) at the
end of 1928 and the percent of a county in high and medium erosion areas. For ease of
45
Subsequent banking is potentially an outcome of the Dust Bowl. The results are similar when using the
amount of deposits prior to the Dust Bowl.
interpretation, the log number of banks is normalized to have a mean of zero and a standard
deviation of one. The specification also controls for the main effect from banks, and an
interaction between banks and each pre-Dust Bowl outcome level to allow for differential
pre-trends by banking levels. The coefficients on these terms are allowed to vary over each
time period.
(3)
Yct - Yc1930
=
ltMc
2tHc
+ 3tBe
+ 04tBc x Mc + 05tBe x He
+
o st + OtXc + y71Lc +
The coefficients of interest are 04t and /5t.
72tB,
X Lc + ct.
These indicate whether counties with more banks
adjusted differently over each time period than similarly eroded counties with fewer banks.
The main identification assumption is that counties with more original banks had better
access to credit during and after the Dust Bowl, but would otherwise have adjusted agricultural land-use similarly. Pre-Dust Bowl banking levels may be correlated with other county
characteristics (credit demand, education levels, overall financial development), so it must
be assumed that these other characteristics do not predict differential responses to erosion.
Table 5, columns 1 to 3, presents the results. More-eroded counties with more banks
immediately shifted toward greater pasture (column 1).46 By contrast, there was no relative
shift from wheat to hay (column 2).
This could reflect increases in animal production
requiring more credit than increases in hay production, due to large capital purchases of
livestock. Estimated changes in total farmland are mixed (column 3). High erosion counties
with more banks had greater declines in farmland over time, but medium erosion counties
experienced some relative increases in farmland.
It is surprising that pre-1930's banking levels would continue to predict differential adjustment into the 1960's, presumably after the recovery of local credit markets. There may
46
When estimating equation (4) without the interaction between banking and pre-Dust Bowl land allocation, there is no pre-trend toward greater pasture in more-eroded counties with more banks.
be lock-in effects once early adjustments differ, but the interaction effect is strengthening
slightly. This highlights the possibility that banking levels may be correlated with other
characteristics that predict adjustment to erosion.
Low Incentives for Tenants to Adjust. When land is rented with imperfect contracting, tenants may have inefficiently low incentives to make permanent investments in the
land. Contemporaries emphasized that tenants' focus was on their crop rather than the land
(McDonald 1938). Thus, the presence of tenant farmers may have reduced adjustments to
restore land productivity or prevent future erosion. 47
This hypothesis is tested by estimating equation (4), but replacing the number of banks
with the share of farmland managed by tenants in 1930 (normalized to have a mean of
zero and a standard deviation of one). Table 5, columns 4 to 6, present the results on
the interaction term. There are few systematic patterns in the data, though there is some
evidence of counties with less tenant farmland switching more from wheat to hay, which
would also be less likely to worsen erosion.
For a narrow window around 1930 to 1950, there is data on tenants' land value, cropland,
and equipment. Comparing counties with different erosion levels, tenants and non-tenants
are estimated to have had similar changes in the per-acre value of farmland, the share of
cropland in total farmland, and per-acre equipment. These estimates suggest that tenants
may have had similar adjustment incentives.
If tenant contracts give inefficient adjustment incentives, then overall farming should
adjust away from land being managed by tenants. Figure 11 presents the results from
estimating equation (1) for the share of land farmed by tenants. Tenant shares did not
decline immediately, but there is some indication of declines in the 1950's and after. This
shift away from tenancy, along with the move toward larger farm sizes, could indicate that
long-run adjustment in land-use required the reorganization of land ownership.
47
Tenant farmers could also be more credit constrained, but this is not obvious: land owners would have
lost substantial capital from lower land values, while tenants' assets might have been more liquid and allowed
greater adjustment.
Government Payment Programs. There was a substantial increase in government
payment programs during the 1930's, and many programs were targeted at the agricultural
sector. To the extent that these payments were also targeted to higher erosion counties, they
could have distorted farmers' incentives to adjust agricultural production. The New Deal
agricultural payment programs tended to encourage the analyzed production adjustments,
as well as the general continuation of farming.
Table 6, panel A, presents information on average payments for different New Deal programs, and the correlation between payments and county erosion levels, after controlling for
state fixed effects and the usual county characteristics. There is little systematic relationship
between program payments and erosion levels. It is possible, however, that farmers expected
future payments.
Parts of the New Deal programs were temporary, but some parts were extended or
changed into other forms of agricultural payments. Panel B of table 6 presents information on all government payments, beginning again in 1969. Higher erosion counties did not
receive more payments in 1969, but began to in 1974. Total agricultural payments were
much higher in 1987, but higher erosion counties received substantially less. This is likely
due to the 1985 introduction of the Conservation Reserve Program (CRP), which paid farmers to take low-quality and erosion-prone land out of production. This precluded other farm
payments. In 1992 and 1997, the higher erosion counties receive more CRP payments.
Overall, government payments were not targeted to more-eroded counties, so these programs seem unlikely to have discouraged agricultural adjustment to the Dust Bowl erosion.
However, these estimates are consistent with a persistent drop in land quality. In the 1990's,
a higher fraction of government payments were through the CRP, which targets low-quality
erodible lands. These lands were still disproportionally in those counties more eroded during
the Dust Bowl.
V.E
Adjustment in Population, Industry, and Wages
In response to the substantial costs of the Dust Bowl, there should be strong incentives to
reallocate labor across locations and/or sectors. Given that the agricultural sector was slow
to adjust, this may create extra pressure for adjustment along the available margin of labor.
However, if the agricultural sector was slow to adjust away from labor-intensive activities,
this persistence could slow the reallocation of labor.
Population. Panel A of figure 12 graphs changes in log population from estimating
equation (1). Prior to the 1930's, population was increasing in high erosion counties relative
to low erosion counties. In medium erosion counties, population increased relatively from
1910 to 1920 and then decreased from 1920 to 1930. Following the Dust Bowl, population
decreased substantially in both high and medium erosion counties.
In order to control for pre-1930 differences in levels and trends, column 1 of table 7
presents the numerical results from estimating equation (2). During the 1930's, high erosion
counties experienced an 8.1% decrease in population, relative to low erosion counties in the
same state. Medium erosion counties had a 6.5% relative decrease in population. In both
county types, population continued to decline through the 1950's. By comparison, state-wide
populations decreased 3% - 4% in Oklahoma, Kansas, and Nebraska from 1930 to 1940.
To explore selection in which populations decreased, panel B of figure 12 graphs changes
in the fraction of population in rural areas of the county (areas with fewer than 2,500 inhabitants) and column 2 of table 7 presents the numerical results. The fraction of population
in rural areas was mostly unchanged, indicating that decreases in population did not occur
predominately in rural areas. Similarly, estimates in column 3 of table 7 show that the decrease in population was not especially among those living on farms. 48 There was sometimes
a smaller decrease among farm populations, which may otherwise be less likely to migrate.
The widespread population decrease in these counties may also indicate important spillover
effects between agriculture and industry, perhaps through locally supplied inputs.
48
Data on farm populations are not available before 1930.
Historical survey data provide some evidence on the selection of Dust Bowl migrants.
Janow and McEntire (1940) describe results from a special survey of 1930's migrants to
California, many of whom came from Oklahoma. 49 Comparing the distribution of migrants'
original occupations in Oklahoma to the distribution of all Oklahoma residents occupations
in 1930, there is only a small tendency for migration from the agricultural sector. 50 By
contrast, migrants from other regions were less likely to be from the agricultural sector. The
Dust Bowl may have shifted the typical selection of migrants toward the agricultural sector,
but not much beyond proportional levels.
Unemployment and Manufacturing Employment. From 1930 to 1940, the overall
unemployment rate increased by 0.71 percentage points in high erosion counties (column 1,
table 8), and this increase was gone by 1950. Medium erosion counties had no relative increase in unemployment by 1940. These estimates suggest that surplus labor was reallocated
relatively quickly.
For a restricted sample of counties, it is possible to analyze within-county shifts in labor
toward the manufacturing sector. I focus on the manufacturing sector, instead of wholesale
and retail, because it would tend to be less directly affected by local changes in demand.
However, there may be important spillovers between agriculture and manufacturing, as suggested by the widespread decrease in population.
Column 2 of table 8 reports changes in the fraction of the labor force employed in manufacturing.51 Column 3 reports changes in the fraction of the population employed in manufacturing. In high erosion counties, there were small increases from 1930 to 1940 in the
49
The survey was conducted by the Bureau of Agricultural Economics, in cooperation with the California
Department of Education. They surveyed families with school children that had migrated to California in
the 1930's. This survey covered 116,000 families with 187,000 school children, which the surveyors estimated
to be 84% of all such families. The results are described by Janow and McEntire (1940). I have not found
the original data, but have converted the published figures back to aggregated numerical data when possible.
SoComparing the fractions in each occupation for migrant male heads of families vs. all employed males:
31% vs. 28% farmer; 19% vs. 15% farm laborer; 13% vs. 6% semi-skilled laborer; 12% vs. 12% skilled laborers;
12% vs. 15% other laborer; 5% vs. 10% clerks; 5% vs. 8% owner/manager; 1% vs. 2% servant; 0.5% vs. 3%
professional.
51
The labor force is defined as all employed workers, laidoff workers, and unemployed workers searching
for a job.
proportion of workers or population employed in manufacturing, though these are not statistically significant.
These estimates represent large percent increases in manufacturing
employment of 11% and 15%, but they do not account for much overall movement of labor
because manufacturing was a small sector of the economy. In medium erosion counties, there
was no immediate shift in labor, but there were some later increases.
Columns 4 and 5 of table 8 report that total manufacturing establishments and value
added did not increase following the Dust Bowl, though the coefficients are imprecisely estimated. Manufacturing may have been too slow to expand, perhaps due to the Depression, to
attract workers before they left the county. Even after the Depression, there was no increase
in manufacturing and the reallocation of labor continued through population declines. This
overall pattern of results is similar to that found by Blanchard and Katz (1992) for state-level
responses to a labor demand shock in the second half of the
2 0 th
century.
Proxies for Wages. Completing the empirical picture of labor market adjustment requires knowledge of changes in wage rates. Micro data on changes in wages are not available,
but per-capita retail sales may serve as a proxy for local income (Fishback et al. 2005). This
measure would differ from wages if there is net savings or borrowing, and would more closely
reflect income if individuals' labor supply changed.
Column 1 of table 9 reports that high erosion and medium erosion counties experienced
a 9.8% and 6.3% decrease in per-capita retail sales from 1930 to 1940. If this were the
decrease in total income, it would be partly offset by lower land prices. A rough calculation
indicates that individuals spent about 1% of their budget on land for housing. 52 Decreased
land values would then compensate workers for a decrease of 2.78% and 1.67% in high and
medium erosion counties, respectively.
Per-capita retail sales recovered from 1940 to 1958, partially in high erosion counties and
completely in medium erosion counties. If we assume that income or wages recovered by a
52
Assume that annual rental rates on residential land are 7% of the average value of one acre of agricultural
land. The budget share is found by dividing this number by average per-capita retail sales. The number is
similar whether unweighted or weighting by farmland or population.
similar percent, it is useful to check whether the magnitude of recovery would require relative
changes in labor demand or if it could be explained by the observed decreases in population
over the same time period. Reinterpreting estimates by Borjas (2003) on immigration would
imply that a 10% decrease in population would increase wages by 3% to 4%, or increase
income by 6.4% if individuals' labor supply is allowed to increase. In high erosion counties,
the population decrease from 1940 to 1960 implies a 2% to 2.7% increase in wages and a
4.3% increase in income - compared to an observed 4% recovery in per-capita retail sales
from 1940 to 1958. In medium erosion counties, the population decrease implies a 3.2% to
4.2% increase in wages and a 6.7% increase in income - compared to a 6.8% recovery in
per-capita retail sales.
For a restricted sample of counties, additional data is available on total payroll and
workers in the manufacturing, wholesale, and retail sectors. Columns 2 to 4 of table 9 report
estimated changes in the log of payroll divided by workers, as a proxy for salary income. In
the retail sector, for which most counties are available, there is an insignificant immediate
decline in salary and the long-run declines are similar or more negative than for per-capita
retail sales. In the wholesale and manufacturing sectors, there are greater long-run declines in
salaries. Given the large decreases in population, this may reflect changes in the underlying
skill distribution of the workforce. Otherwise, relative wages are at their lowest only when
relative changes in population cease.
Interactions with Agriculture.
Changes in the labor market potentially interact
with adjustment incentives in the agricultural sector. There is some evidence that wages
decreased in more-eroded counties, and this might be especially true in the agricultural
sector if switching sectors is costly.
Consistent with temporarily lower wages, estimates indicate that the ratio of labor to
capital machinery increased temporarily in the agricultural sector. This ratio is approximated by dividing the number of people living on farms by the value of equipment and
machinery on farms. Changes in the log of this ratio are estimated by equation (2). For
high erosion counties, the changes (and standard error) after 1930 are: 14% (4.8%) by 1940,
2% (4.8%) by 1945, and 9% (5.5%) by 1969. For medium erosion counties, the changes after
1930 are: 13% (4.1%) by 1940, 4.5% (4.1%) by 1945, and 2.4% (4.3%) by 1969.
Despite the immediate decreases in population, there may still have been surplus labor
in 1940. This is reflected in somewhat higher unemployment and the subsequent decreases
in population. Some of this surplus labor appears to have been absorbed in the agricultural
sector. This may have discouraged switching land from crops to animals or from wheat to
hay, which use less labor. It was in the 1950's, as population declines ceased, that agricultural
adjustment began to appear along these margins.
VI
Conclusion
The 1930's American Dust Bowl provides a unique context in which to examine economic
adjustment to a permanent change in the environment. Dust Bowl erosion imposed substantial short-run costs on Plains agriculture. Adjustment in agricultural land-use was slow and
limited, despite evidence of potentially productive adjustments. Estimates imply that the
long-run cost to agriculture was 86% of the short-run cost.
In other settings where short-run production costs from a shock are substantial, slow
and limited adjustment would imply that the long-run production costs may also be large.
This paper provides some evidence on the role of credit and land ownership in predicting
adjustment, and an important area for further research is understanding what conditions
facilitate long-run adjustment. Further research on historical shocks may help anticipate the
long-run effects of modern shocks for which only short-run effects can be observed.
If long-run production costs are large, the overall welfare consequences may be partly
lessened if people can leave the area. The Unites States is known to have high labor mobility,
so migration may be more limited in response to shocks in other countries. In particular,
cross-country migration may be severely restricted by political constraints. If short-run
production costs persist and migration is limited, shocks that have large short-run costs may
impose similarly large long-run costs.
References
Allred, B.W. and W.M. Nixon. 1955. "Grass for Conservation in the Southern Great Plains." Farmer's
Bulletin No. 2093.
Ashcraft, Adam. 2005. "Are Banks Really Special? New Evidence from FDIC-induced Failures of Healthy
Banks." American Economic Review, Vol. 95, No. 5, pp. 1712-1730.
Bernanke, Ben S. 1983. "Nonmonetary Effects of the Financial Crisis in the Propagation of the Great
Depression." American Economic Review, Vol. 73, No. 3, pp. 257-276.
Blanchard, Olivier Jean, and Lawrence F. Katz. 1992. "Regional Evolutions." Brookings Paperson
Economic Activity, Vol. 1992, No. 1, pp. 1-61.
Bloom, Nick, Stephen Bond, and John Van Reenen. 2007. "Uncertainty and Investment Dynamics." Review
of Economic Studies, Vol. 74, pp. 391-415.
Borjas, George J. 2003. "The Labor Demand Curve is Downward Sloping: Reexamining the Impact of
Immigration on the Labor Market." QuarterlyJournalofEconomics, Vol 118, No. 4, pp. 1335-1374.
Boustan, Leah Platt, Price V. Fishback, and Shawn E. Kantor. 2007. "The Effect of Internal Migration on
Local Labor Markets: American Cities During the Great Depression." NBER WP#13276.
Calomiris, Charles W., and Joseph R. Mason. 2003. "Consequences of Bank Distress during the Great
Depression." American Economic Review, Vol. 93, No. 3, pp. 937-947.
Carrington, William J. 1996. "The Alaskan Labor Market during the Pipeline Era." Journal of Political
Economy, Vol. 104, No. 1, pp. 186-218.
Chari, V. V. and Hugo Hopenhayn. 1991. "Vintage Human Capital, Growth, and the Diffusion of New
Technology." Journalof PoliticalEconomy, Vol. 99, No. 6, pp. 1142-1165.
Chepil, W.S., F. H. Siddoway, and Dean V. Armbrust. 1962. "Climatic Factor for Estimating Wind
Erodibility of Farm Fields." Journalof Soil and Water Conservation,Vol. 17, pp. 162 - 165.
Conley, T.G. 1999. "GMM estimation with cross sectional dependence." Journalof Econometrics, Vol. 92,
pp. 1-45.
Cunfer, Geoff. 2005. On the Great Plains: Agriculture and Environment. Texas A&M University Press,
College Station.
Davis, Donald R., and David E. Weinstein. 2002. "Bones, Bombs, and Break Points: The Geography of
Economic Activity." American Economic Review, Vol. 92, No. 5, pp. 1269-1289.
Dell, Melissa, Benjamin F. Jones, and Benjamin A. Olken. 2008. "Climate Change and Economic Growth:
Evidence from the Last Half Century." NBER WP#14132.
Deschenes, Olivier, and Michael Greenstone. 2007. "The Economic Impacts of Climate Change: Evidence
from Agricultural Output and Random Fluctuations in Weather." American Economic Review, Vol. 97,
No. 1, pp. 354-385.
Diamond, Jared. 2005. Collapse: How Societies Choose to Fail or Succeed.
Dixit, A. and R. Pindyck. 1994. Investment Under Uncertainty. Princeton University Press. Princeton, New
Jersey.
Duflo, Esther. 2004. "The medium run effects of educational expansion: evidence from a large school
construction program in Indonesia." Journalof Development Economics, Vol. 74, pp. 163-197.
Fishback, Price V., William Horrace, and Shawn Kantor. 2005. "Did New Deal Grant Programs Stimulate
Local Economies? A Study of Federal Grants and Retail Sales During the Great Depression." Journalof
Economic History, Vol. 65, No. 1, pp. 36-71.
Foster, Andrew D., and Mark R. Rosenzweig. 1995. "Learning by Doing and Learning from Others: Human
Capital and Technical Change in Agriculture." Journalof PoliticalEconomy, Vol. 103, No. 6, pp. 11761209.
Griliches, Zvi. 1957. "Hybrid Corn: An Exploration in the Economics of Technical Change." Econometrica,
Vol. 25, No. 4, pp. 501-522.
Guiteras, Raymond. 2008. "The Impact of Climate Change on Indian Agriculture." Mimeo.
Hansen, Zeynep K. 2005. "Book Review of On the Great Plains: Agriculture and Environment by Geoff
Cunfer." Journal ofEconomic History, Vol. 65, No. 4, pp. 1162-1163.
Hansen, Zeynep K. and Gary D. Libecap. 2004. "Small Farms, Externalities, and the Dust Bowl of the
1930s." Journalof PoliticalEconomy, Vol. 112, No. 3, pp. 665 - 694.
Hornbeck, Richard. 2008. "Good Fences Make Good Neighbors: Evidence on the Effects of Property
Rights." Mimeo.
Hoyt, John C. 1936. "Droughts of 1930 - 1934." Water-Supply Paper 680, U.S. GPO.
Hurt, Douglas R. 1981. The Dust Bowl: An Agricultural and Social History. Nelson Hall, Chicago.
Hurt, Douglas R. 1985. "The National Grasslands: Origin and Development in the Dust Bowl," in The
History of Soil and Water Conservation, eds. Douglas Helms and Susan L. Flader. The Agricultural
History Society, Washington D.C., pp. 144 - 157.
Janow, Seymour J., and Davis McEntire. 1940. "The Migrants: Migration to California." Land Policy
Review, July-August, pp. 24-36.
Leuchtenburg, William E. 1963. Franklin D. Roosevelt and the New Deal, 1932 - 1940. Harper & Row,
Publishers, New York.
Margo, Robert A. 1997. "Wages in California during the Gold Rush." NBER WP on Historical Factors in
Long Run Growth, Historical Paper 101.
McDonald, Angus. 1938. "Erosion and its Control in Oklahoma Territory." Soil Conservation Service,
Miscellaneous Publication No. 301, GPO.
Miguel, Edward, and Gerard Roland. 2006. "The Long Run Impact of Bombing Vietnam." Mimeo.
National Archives (College Park, MD), Record Group 114, Cartographic Records of the Soil Conservation
Service, Entry #149.
News-week. 1936. "Drought: A Merciless Sun and a Scourge of Insects Destroy Crops, Cattle and Men Two-thirds of the Country Afflicted." July 18.
Phillips, Sarah T. 2007. This Land, This Nation: Conservation, Rural America, and the New Deal.
Cambridge University Press.
Pressly, Thomas J. and William H. Scofield. 1965. Farm Real Estate Values in the United States by Counties,
1850 - 1959. University of Washington Press, Seattle.
Rasmussen, Wayne D. and Gladys L. Baker. "Price-Support and Adjustment Programs From 1933 Though
1978: A Short History." Agricultural Information Bulletin No. 424, U.S. Department of Agriculture,
U.S. Government Printing Office.
Redding, Stephen J., and Daniel M. Sturm. Forthcoming. "The Costs of Remoteness: Evidence from German
Division and Reunification." American Economic Review.
Roback, Jennifer. 1982. "Wages, Rents, and the Quality of Life." Journalof PoliticalEconomy, Vol. 90, No.
6,pp. 1257 - 1278.
Rosenzweig, Mark R., and Kenneth I.Wolpin. 2000. "Natural 'Natural Experiments' inEconomics." Journal
of Economic Literature, Vol. 38, No.4,pp. 827-874.
Saloutos, Theodore. 1982. The American Farmer and the New Deal. Iowa State University Press.
Science. 1934. "The Recent Destructive Dust Cloud." New Series, Vol. 79, May 25, p.473.
Soil Conservation Service. 1935. "Soil Erosion: A Critical Problem in American Agriculture." Part V of the
Supplementary Report of the Land Planning Committee to the National Resources Board. US GPO.
Soil Conservation Service. 1937. "Soil Conservation Districts for Erosion Control." Miscellaneous
Publication No. 293, GPO.
Soil Conservation Service. 1939. "Land Facts on the Southern Plains." Miscellaneous Publication No. 334,
GPO.
Soil Conservation Service. 1955. "Facts About Wind Erosion and Dust Storms on the Great Plains." Leaflet
No. 394. US Department of Agriculture.
US Department of Agriculture. 1940. "The Dust Bowl: Agricultural Problems and Solutions," Editorial
Reference Series, No. 7, July 15.
Wallace, Henry A. 1938. "Forward" in Soils and Men: Yearbook of Agriculture. US Department of
Agriculture, Washington D.C., Government Printing Office.
Worster, Donald. 1979. Dust Bowl: The Southern Plains in the 1930s. Oxford University Press, Oxford.
Figure 1. Aggregate Changes on the Plains in Agriculture and Population
A. Farmland Value, per farm acre (Log/CPI)
B. Agricultural Revenue, per farm acre (Log/PPI)
,
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
1980
1990
2000
I
C. Farmland Acres, per county acre
(0
Ii
1910
1920
1930
1940
1950
1960
1970
E. Fallow Acres, per cropland acre
1910
1920
1930
1940
1950
1960
1970
F. Average Farm Size (Log)
1980
1990
1910
2000
1920
1930
1940
1950
1960
1970
1980
1990
2000
1980
1990
2000
H. %Population in Rural Areas
G. Population (Log)
,
H. % Population in Rural Areas
U1
2
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
1910
1920
1930
1940
1950
1960
1970
Notes: All panels (except E) report values aggregated over the same 769 Plains counties in every period for the following outcomes:
Panel A, the log per-acre value of farmland divided by the consumer price index; Panel B, the log per-acre agricultural revenues divided
by the producer price index for farm products; Panel C, the fraction of county land in farms; Panel D, the fraction of farmland that is
cropland; Panel E, the fraction of cropland that is fallow (available for only 587 counties); Panel F, the log acres of farmland divided by
the number of farms; Panel G, the log population; Panel H, the fraction of population living in areas with fewer than 2,500 inhabitants.
49
Figure 2. Sample Counties, Shaded by Erosion Level
Low Erosion
Medium Erosion
High Erosion
Not In Sample
Notes: Mapped erosion levels are indicated as low (less than 25% of topsoil lost), medium (25%-75% of topsoil lost and may have
some gullies), or high (more than 75% of topsoil lost and may have numerous or deep gullies). Thin lines denote 1910 county borders,
corresponding to the sample of 769 counties described in Table 1. Thick lines denote state boundaries. Crossed out areas are not in the
sample.
Figure 3. Log Changes in the per-Acre Value of Agricultural Land, by Erosion Level
1910
1920
1930
1940
----
High vs. Low
1950
1960
-----
1970
1980
1990
2000
Medium vs. Low
Notes: This figure graphs the estimated coefficients (P) from equation (1) in the text. For each year, the solid circles report differences
in the log per-acre value of farmland for high erosion counties, relative to low erosion counties. The hollow squares report differences
for medium erosion counties, relative to low erosion counties. These coefficients are estimated by regressing the per-acre value of
farmland on the percent of a county in a high erosion area (solid circle) and the percent of a county in a medium erosion area (hollow
square), controlling also for state-by-year fixed effects, the interaction between each year and each county characteristic in panels B - E
of table 1, and the interaction between each year and the available lagged values of each county characteristic in panels B - E of table 1.
Figure 4. Log Changes in Agricultural Revenue and Capital, by Erosion Level
A. Agricultural Revenue, per Acre of Farmland
C'
/
cr
Ci
1910
1920
1930
1940
-
1960
1950
High vs. Low
--
&--
1970
1980
1990
2000
Medium vs Low
B. Value of Capital (Machinery & Equipment), per Acre of Farmland
--
-----------
a
I
L
1910
-
1920
1930
1940
---
High vs Low
1950
1960
-----
1970
1980
1990
2000
Medium vs. Low
Notes: For the indicated outcome variable, each panel graphs the estimated coefficients (0) from equation (1) in the text, as described in
the notes to Figure 3.
Figure 5. Log Changes in Land Value, Plains Counties vs. Non-Plains Counties
A. High - Low Erosion
U-
,
,
1910
1920
|
193 0
,
,
,
I
1940
1950
1960
1970
- Plains
--
----
1980
1990
2000
1980
1990
2000
Non-Plains
B. Medium - Low Erosion
C.,
Ci
-----------
1910
1920
1930
1940
---
1950
Plains
1960
----
1970
Non-Plains
Notes: The estimates in both panels are from a single regression of equation (1) in the text, modified to include only state-by-time fixed
effects as controls. Panel A reports the difference between high and low erosion counties, for both Plains and non-Plains states Panel
B reports the difference between medium and low erosion counties, for both Plains and non-Plains states. Plains states are those
depicted in Figure 2. The sample includes 867 Plains counties and 1840 non-Plains counties.
Figure 6. Changes in Farmland Acres per county acre, by Erosion Level
0-
LO)
1910
1920
1930
1940
1950
---
High vs. Low
1960
----
1970
1980
1990
2000
Medium vs. Low
Notes. This figure graphs the estimated coefficients (3) from equation (1) in the text, as described in the notes to Figure 3. Omitted as
controls are the 1930 and lagged values of farmland per county acre.
Figure 7. Log Changes in Crop & Animal Productivity, and Changes in Cropland as a Fraction of Cropland and Pasture
A. Crop and Animal Productivity: High - Low Erosion
B. Crop and Animal Productivity: Medium - Low Erosion
(N
0
(N
1910
1920
1930
-
1940
1950
1960
Crop Productivity
1970
1980
1990
2000
1910
1920
Animal Productivity
1930
1940
1950
- Crop Productivity
I.Allocation of Land to Crops: High - Low Erosion
1960
a
1970
1980
1990
2000
Animal Productivity
D. Allocation of Land to Crops: Medium - Low Erosion
to
o
C
I
i
N
In
N
to
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
Notes: Panels A and C graph coefficients from equation (1), estimated separately for crop and animal productivity. Panel A reports the change in high erosion relative to low erosion,
Panel B reports changes in medium relative to low erosion (normalized to zero in 1930). Panels C and D graph estimated changes in the relative fraction of land allocated to crops.
Figure 8. Log Changes in Wheat & Hay Productivity, and Changes in Wheat Land as a Fraction of Land in Wheat and Hay
B. Wheat and Hay Productivity: Medium - Low Erosion
A. Wheat and Hay Productivity: High - Low Erosion
1910
1920
1930
-
-
1940
1950
Wheat Productivity
1960
1970
1980
----
Hay Productivity
1990
-1910
1910
2000
1920
1930
-
1940
1950
1960
Wheat Productivity
1970
1980
1990
2000
Hay Productivity
D. Allocation of Land to Wheat: Medium - Low Erosion
C. Allocation of Land to Wheat: High - Low Erosion
I
0
i
0I
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
1910
1920
1930
1940
1950
1960
1970
1980
1990
2000
Notes: Panels A and C graph coefficients from equation (1), estimated separately for wheat and hay productivity. Panel A reports the change in high erosion relative to low erosion,
Panel B reports changes in medium relative to low erosion (normalized to zero in 1930). Panels C and D graph estimated changes in the relative fraction of land allocated to wheat.
Figure 9. Changes in Average Farm Size and Land Fallowing, by Erosion Level
A. Log Average Farm Size
A
"
-"--
,
,
!
.
.
.
.
.
1920
19'30
1940
1950
1960
1970
1980
---
High vs. Low
'U-
\
, " "
1910
-- e--- Medium vs. Low
a
1990
2000
1990
2000
I
B. Acres of Fallow Land, per acre of cropland
9
1910
1920
1930
-
1940
1950
High vs. Low
1960
--
E--
1970
1980
Medium vs. Low
Notes: For the indicated outcome variable, each panel graphs the estimated coefficients (13)from equation (1) in the text, as described in
the notes to Figure 3. Panel A omits 1930 and lagged values of average farm size as a control variable.
Figure 10. Log Changes in the Number of Active Banks, by Erosion Level
"
-- e
..a.
-
1920
1925
1920
---
High vs. Low
1929
-----
1933
1936
Medium vs. Low
Notes: This figure graphs the estimated coefficients (13)from a modified version of equation (1), described in the notes to Figure 3.
Banking data is available annually, so the regression controls for state-by-year fixed effects and county characteristics interacted with
every year.
Figure 11. Changes in the Tenant Share of Farmland, by Erosion Level
LO
I
1910
1920
BI
1930
--
1940
-
1950
High vs. Low
1960
-- E--
1970
1980
1990
2000
Medium vs. Low
Notes: This figure graphs the estimated coefficients (13)
from equation (1) in the text, as described in the notes to Figure 3.
Figure 12. Changes in Population and Rural Population, by Erosion Level
A. Log Population
i\
--------
1910
1920
1930
---
1940
1950
High vs. Low
1960
--
1970
1980
1990
2000
B-- Medium vs. Low
B. Fraction of Population in Rural Areas of County
0
s---_
-13----a-
1910
1920
1930
----
1940
1950
High vs Low
1960
--
e---
1970
1980
1990
2000
Medium vs. Low
Notes: For the indicated outcome variable, each panel graphs the estimated coefficients (P) from equation (1) in the text, as described in
the notes to Figure 3. In panel A, the 1930 and lagged values of population per acre are omitted as controls. In panel B, omitted controls
are the 1930 and lagged values of fraction of population in rural areas.
Table 1. County Characteristics in 1930, by Post-Dust Bowl Erosion Level
All Counties
(1)
Panel A: Value
Value of farm land & buildings,
per acre in farms
Value of farm land,
per acre in farms
Value of all farm products,
per acre in farms
Panel B: Land-use
Acres of land in farms,
per county acre
Acres of cropland,
per acre in farms
Panel C: Population and Farms
Population, per 100 county acres
Percent of population, rural areas
Percent of population, on farms
Farms, per 100 county acres
Average farm size (acres)
Panel D: Cropland Allocation
% Corn
% Wheat
% Hay
% Cotton
% Oats, Barley, and Rye
Panel E: Animal Production
Cows, per county acre
Relative to Low Erosion:
High Erosion
Medium Erosion
(3)
(2)
(3) - (2)
(4)
3.527
[0.853]
3.309
[0.819]
2.107
[0.789]
0.244*
(0.114)
0.217
(0.111)
0.258**
(0.100)
0.225
(0.154)
0.202
(0.150)
0.202
(0.132)
-0.019
(0.106)
-0.014
(0.103)
-0.055
(0.100)
0.819
[0.187]
0.491
[0.229]
0.035
(0.019)
0.035
(0.031)
-0.004
(0.023)
-0.008
(0.040)
-0.039
(0.020)
-0.043
(0.028)
3.524
[7.018]
0.818
[0.221]
0.542
[0.175]
0.328
[0.246]
613
[1145]
1.289
(0.824)
-0.022
(0.030)
0.020
(0.024)
0.163**
(0.026)
-233
(184)
1.262
(0.826)
0.034
(0.043)
0.040
(0.032)
0.148**
(0.031)
-307*
(152)
-0.027
(0.773)
0.057
(0.042)
0.020
(0.032)
-0.015
(0.031)
-74
(142)
0.188
[0.166]
0.181
[0.214]
0.157
[0.177]
0.104
[0.209]
0.134
[0.116]
0.070**
(0.017)
-0.043
(0.027)
-0.044
(0.030)
0.053**
(0.017)
-0.005
(0.009)
0.192**
(0.027)
-0.102**
(0.035)
-0.099*
(0.044)
0.020
(0.019)
-0.033*
(0.013)
0.122**
(0.023)
-0.059
(0.034)
-0.056*
(0.022)
-0.033
(0.019)
-0.028
(0.010)
0.003
0.013**
0.011**
0.056
(0.003)
(0.004)
(0.003)
[0.033]
0.015
0.066
0.040**
0.055**
Pigs, per county acre
(0.011)
(0.012)
(0.008)
[0.096]
0.001
0.115**
0.116**
0.296
Chickens, per county acre
(0.034)
(0.031)
[0.279]
(0.025)
acres
of
farmland in 1930
are
weighted
by
the
counties
counties,
where
values
for
the
769
sample
Column
1
presents
average
Notes:
(counties based on 1910 borders; with at least 1000 acres of farmland in every period; and with data on each variable in Figure 1 for each
period shown). The standard deviation is reported in brackets.
Columns 2 and 3 report coefficients from a single equation that regresses the indicated county characteristic in 1930 on the
percentage of the county in a medium erosion area and in a high erosion area (low erosion is the omitted category), conditional on State
fixed effects and weighted by acres of farmland in 1930. Column 4 reports the difference between the coefficients in columns 2 and 3.
Robust standard errors are reported in parentheses. ** denotes statistical significance at the 1%level, * at the 5% level.
Table 2. Log Changes in Land Value and Revenue per farm acre, by Erosion Level
Land Value
% Persisting
Change
After 1930
After 1940
(2)
(1)
Panel A. 1940
High - Low
Medium - Low
High - Medium (calculated)
Revenue
Change
% Persisting
After 1940
After 1930
(4)
(3)
0.881
(0.120)
0.824
(0.136)
0.981
(0.313)
0.861
(0.112)
-0.316
(0.055)
-0.202
(0.039)
-0.114
(0.048)
-0.278
(0.041)
-0.167
(0.029)
-0.111
(0.032)
Averaged Value (GLS)
Panel B. 1945
High - Low
Medium - Low
High - Medium (calculated)
-0.234
(0.037)
-0.101
(0.027)
-0.133
(0.031)
0.841
(0.087)
0.609
(0.105)
1.189
(0.238)
0.765
(0.080)
-0.122
(0.044)
-0.131
(0.031)
0.009
(0.035)
0.386
(0.118)
0.648
(0.158)
-0.081
(0.324)
0.452
(0.114)
-0.240
(0.040)
-0.118
(0.027)
-0.123
(0.033)
0.864
(0.127)
0.705
(0.129)
1.101
(0.307)
0.788
(0.112)
-0.208
(0.051)
-0.183
(0.036)
-0.025
(0.040)
0.658
(0.140)
0.904
(0.194)
0.220
(0.326)
0.710
(0.135)
-0.315
(0.042)
-0.167
(0.033)
-0.148
(0.036)
1.133
(0.161)
1.001
(0.196)
1.331
(0.401)
1.094
(0.150)
-0.255
(0.063)
-0.130
(0.044)
-0.125
(0.050)
0.807
(0.193)
0.643
(0.217)
1.099
(0.531)
0.746
(0.178)
-0.298
(0.057)
-0.100
(0.045)
-0.198
(0.051)
1.072
(0.197)
0.603
(0.249)
1.773
(0.557)
0.935
(0.184)
-0.323
(0.092)
-0.056
(0.066)
-0.267
(0.076)
1.021
(0.308)
0.276
(0.321)
2.347
(1.110)
0.693
(0.272)
Averaged Value (GLS)
Panel C. 1950 - 1954 (pooled)
High - Low
Medium - Low
High - Medium (calculated)
Averaged Value (GLS)
Panel D. 1959 - 1969 (pooled)
High - Low
Medium - Low
High - Medium (calculated)
Averaged Value (GLS)
Panel E. 1978 - 1992 (pooled)
High - Low
Medium - Low
High - Medium (calculated)
Averaged Value (GLS)
Ratio of Change
(1)/(3)
(5)
0.8925
0.9496
R-squared
Sample Counties
769
769
Notes: Columns 1 and 3 report the results from estimating equation (2) in the text, for the value of land and revenue per-acre of
farmland, weighting by county farmland in 1930. This estimates the change after 1930, controlling for the same variables as in Figure 3
as well as the values of both land values and revenue in 1930, 1925, 1920, and 1910. The indicated multiple year periods are pooled,
such that the estimate is an average over those years. Columns 2 and 4 report the estimated change, as a proportion of the initial change
from 1930 to 1940. Column 5 reports the change in land value as a proportion of the change in revenue from 1930 to 1940. Reported
in parentheses are robust standard errors clustered by county.
Table 3. Instrumental Variables Estimate of the Change in Land Value from 1930 to 1940
First-Stage:
% of County in
% of County in
High Erosion
(1)
Medium Erosion
(2)
2SLS:
Change in Land
OLS:
Change in Land
Value, 1930 - 1940
(3)
Value, 1930 - 1940
(4)
Erosion Level:
High - Low
Medium - Low
Drought Instruments:
Months in Extreme
Drought, 1930s
Months in Severe
Drought, 1930s
Months in Moderate
Drought, 1930s
Average Palmer Drought
Severity Index, 1930s
0.0061**
(0.0021)
0.0031
(0.0016)
-0.0014
(0.0016)
0.0487
(0.0334)
0.0057
(0.0032)
0.0069**
(0.0022)
0.0076**
(0.0025)
0.1700**
(0.0603)
Controls:
Drought Vars, 1895-1929
State Fixed Effects
County Characteristics
YES
YES
YES
YES
YES
YES
-0.631**
(0.219)
-0.125
-0.277**
(0.041)
-0.165**
(0.154)
(0.030)
YES
YES
YES
YES
YES
YES
F-stat: Instruments
4.69
4.54
0.0010
0.0013
P-value: Instruments = 0
766
766
766
766
Sample Counties
Notes: Column 1 reports first-stage estimates from regressing the percent of a county in high erosion areas on the drought instruments,
controlling for drought from 1895 to 1929, state fixed effects, and the same pre-1930's county characteristics as in table 2. Column 2
reports analogous estimates for the percent of a county in medium erosion areas. Column 3 reports two-stage least squares estimates of
the relative change m land values by erosion level, controlling for the same variables as in columns 1 and 2 and instrumenting for
erosion levels using the excluded drought variables. Column 4 reports OLS estimates of the relative change in land value (as in table 2),
controlling for the same variables as in column 3 and for the same sample of counties. All regressions are weighted by county farmland
in 1930. Robust standard errors are reported in parentheses. ** denotes statistical significance at the 1% level, * at the 5% level.
Table 4. Changes After 1930 in Agricultural Production, by Erosion Level
Erosion Level:
High - Low
1940
1945
1950- 1954
1959- 1964
1969- 1974
1978- 1992
County Share
in Farmland
(1)
Log Crop
Productivity
(2)
Crops vs. Animals
Log Animal
Land Share
Productivity
in Crops
(3)
(4)
-0.014
(0.015)
-0.012
(0.017)
-0.032*
(0.014)
-0.025
(0.016)
0.003
(0.016)
-0.022
(0.018)
-0.489**
(0.118)
-0.135
(0.078)
-0.332**
(0.067)
-0.303**
(0.076)
-0.246**
(0.055)
-0.158*
(0.064)
-0.184**
(0.068)
-0.349**
(0.092)
0.009
(0.010)
0.003
(0.011)
-0.007
(0.013)
-0.030*
(0.015)
-0.263*
(0.115)
-0.215**
(0.073)
-0.233**
(0.058)
-0.223**
(0.064)
0.012
(0.060)
-0.035
(0.043)
-0.067
(0.048)
-0.002
(0.010)
0.021
(0.012)
-0.006
(0.010)
-0.015
(0.013)
-0.001
(0.012)
-0.004
(0.013)
-0.464**
(0.086)
-0.305**
(0.064)
-0.310**
(0.054)
-0.272**
(0.066)
-0.099**
(0.038)
-0.103*
(0.042)
-0.095*
(0.046)
-0.111
(0.063)
-0.002
(0.006)
-0.004
(0.007)
-0.012
(0.008)
-0.018
(0.011)
-0.165
(0.091)
-0.102
(0.054)
-0.214**
(0.054)
-0.300**
1997
Medium - Low
1940
1945
1950- 1954
1959- 1964
1969- 1974
1978-1992
1997
Log Wheat
Productivity
(5)
(0.056)
-0.075
(0.046)
-0.096*
(0.033)
-0.128
(0.047)
Wheat vs. Hay
Log Hay
Productivity
(6)
Land Share
in Wheat
(7)
0.027
(0.022)
-0.003
(0.051)
0.090
(0.048)
0.149**
(0.045)
0.199**
(0.046)
0.141*
(0.063)
0.023
(0.020)
0.003
(0.020)
-0.039
(0.022)
-0.067**
(0.025)
-0.110**
(0.032)
-0.016
(0.016)
0.001
(0.047)
0.033
(0.050)
0.002
(0.039)
0.009
(0.042)
0.014
(0.053)
0.005
(0.020)
-0.022
(0.020)
-0.068**
(0.020)
-0.097**
(0.021)
-0.134**
(0.026)
Hansen-Libecap Adjustments
Log Average
Cropland
Farm Size
Fallowed
(8)
(9)
0.052
(0.032)
0.103**
(0.034)
0.065
(0.034)
0.084*
(0.040)
0.081*
(0.041)
0.054
(0.046)
0.019
(0.016)
-0.032*
(0.013)
0.025
(0.021)
0.063*
(0.025)
0.034
(0.028)
0.024
(0.033)
0.014
(0.033)
-0.006
(0.037)
-0.005
(0.010)
-0.050**
(0.010)
-0.044**
(0.011)
-0.034**
(0.011)
-0.018
(0.015)
0.005
(0.016)
R-squared
0.6133
0.7473
0.8133
0.7877
0.8193
0.4607
0.7254
0.8915
0.8094
Sample Counties
769
769
388
769
587
769
769
388
388
Weighted by
Farmland
Farmland
Cropland
Farmland
Pasture
Cropland
Wheat
Wheat
Hay
1930 Value of:
plus Pasture
plus Hay
cc
Notes: For each outcome variable, the column reports estimates from equation (2) in the text and described in notes to Table 2. Where indicated, the weights are adjusted to reflect the
average change for the relative unit of land. Reported in parentheses are robust standard errors clustered by county. ** denotes statistical significance at the 1% level, * at the 5% level.
Table 5. Changes in Land-use after 1930, Interacted with County Pre-Characteristics
Erosion Level:
High - Low
1940
1945
1950 - 1954
1959- 1964
Relative Adjustment in Areas with More Banks
County Share
Land Share
Land Share
in Farmland
in Wheat
in Crops
(2)
(3)
(1)
-0.042**
(0.014)
0.015
0.008
(0.018)
(0.014)
0.025
(0.022)
-0.004
(0.015)
-0.030
(0.018)
-0.039*
-0.050**
(0.017)
-0.071**
(0.019)
-0.074**
(0.020)
1969 - 1974
-0.016
(0.021)
0.000
(0.021)
0.010
(0.023)
Relative Adjustment in Areas with More Tenants
County Share
Land Share
Land Share
in Farmland
in Wheat
in Crops
(6)
(4)
(5)
-0.012
(0.010)
-0.001
(0.012)
-0.013
(0.015)
-0.013
(0.017)
(0.018)
-0.020
(0.018)
0.048*
(0.020)
0.026
(0.021)
0.008
(0.022)
-0.038*
(0.017)
-0.044*
(0.019)
1978 - 1982
1987 - 1992
0.005
(0.017)
-0.004
(0.022)
-0.016
(0.016)
-0.013
(0.018)
-0.035
(0.018)
-0.023
(0.020)
-0.025
(0.022)
Medium - Low
1940
1945
1950 - 1954
1959- 1964
(0.006)
-0.012
(0.012)
-0.024**
(0.007)
-0.013
-0.010
(0.008)
(0.018)
-0.021 *
(0.011)
-0.007
(0.017)
0.001
(0.019)
-0.017**
1969 - 1974
1978- 1982
1987- 1992
0.02 1*
(0.009)
0.007
(0.012)
0.018
(0.010)
0.026*
(0.013)
0.026*
(0.012)
0.015
(0.011)
0.011
(0.013)
-0.006
(0.006)
0.008
(0.008)
-0.009
(0.014)
0.017
(0.010)
0.015
0.056**
(0.018)
0.057**
(0.019)
0.050*
(0.020)
(0.013)
0.007
(0.011)
-0.031 *
(0.015)
0.008
(0.011)
0.027*
(0.014)
0.024
(0.013)
0.033*
(0.013)
0.014
(0.014)
0.7307
0.6087
0.4844
0.4702
0.6180
696
696
696
696
696
Farmland
Farmland
Farmland
Wheat
Wheat
plus Hay
plus Hay
Notes: For the indicated outcome variable, each column presents estimates from equation (3) in the text. For columns 1 - 3, the
reported coefficients are the interaction term: the adjustment to erosion in areas with more banks in 1928, relative to the adjustment in
areas with fewer banks. The log number of banks is normalized to have a mean of zero and a standard deviation of one. Columns 4 - 6
report the analogous coefficients, but for the normalized tenant share of farmland in 1930 (instead of the normalized log number of
banks). Reported in parentheses are robust standard errors clustered by county. ** denotes statistical significance at the 1% level, * at
the 5% level.
R-squared
Sample Counties
Weighted by
1930 Value of:
0.7488
696
Farmland
Table 6. Government Program Payments per farm acre, by Erosion Level
All Counties
(1)
Panel A. New Deal payments (1933-1939)
AAA payments
Public works spending
Relief spending
New deal loans
Mortgage loans guaranteed
Panel B. Government payments
All payments, 1969
0.489
[0.327]
0.264
[0.605]
0.508
[2.435]
0.484
[1.126]
0.112
[0.792]
Relative to Low Erosion:
Medium
High Erosion
Erosion
(3)
(2)
0.001
(0.017)
0.008
(0.058)
0.110
(0.100)
-0.090
(0.094)
0.001
(0.040)
-0.027
(0.023)
-0.033
(0.064)
0.142
(0.129)
-0.087
(0.112)
-0.103
(0.059)
-0.159
-0.245
3.323
(0.284)
(0.256)
[2.646]
0.257**
0.088
0.364
All payments, 1974
(0.066)
(0.056)
[0.390]
-5.511**
-1.838*
16.040
All payments, 1987
(1.076)
[13.631]
(0.866)
-1.323*
-0.887
7.930
All payments, 1992
(0.549)
(0.477)
[5.580]
-0.523
-0.222
8.060
All payments, 1997
(0.467)
(0.366)
[5.843]
1.776**
0.750**
1.571
CRP payments, 1992
(0.263)
[1.490]
(0.162)
2.562**
1.186**
2.094
CRP payments, 1997
(0.350)
[2.061]
(0.225)
0.185**
0.089**
0.217
Fraction of money from CRP, 1992
(0.027)
(0.018)
[0.149]
0.230**
0.105**
0.289
Fraction of money from CRP, 1997
(0.031)
[0.185]
(0.022)
Notes: Panel A reports differences in 1930's New Deal spending across counties, for a constant sample of 766 counties.
(3) - (2)
(4)
-0.028
(0.021)
-0.041
(0.066)
0.032
(0.118)
0.003
(0.087)
-0.104*
(0.042)
0.086
(0.254)
0.169**
(0.045)
-3.673**
(1.075)
-0.436
(0.472)
-0.300
(0.442)
1.026**
(0.245)
1.376**
(0.322)
0.096**
(0.023)
0.125**
(0.027)
Panel B
reports later differences in government payments, conservation reserve program payments, and the fraction of payments through the
conservation reserve program. Column 1 presents the mean for each variable and the standard deviation in brackets. Column 2 reports
the average difference for a county with medium erosion relative to low erosion, controlling for state fixed effects, each characteristic in
Panels B - E of Table 1, and the lagged values of those characteristics. Columns 3 and 4 report the same average differences comparing
high erosion vs. low erosion and high erosion vs. medium erosion, respectively. All variables and regressions are weighted by county
farmland in 1930. Robust standard errors are reported in parentheses. ** denotes statistical significance at the 1% level, * at the 5%
level.
Table 7. Changes After 1930 in Population Outcomes, by Erosion Level
Erosion Level:
High - Low
1940
1950
1960
1970
1990
Medium - Low
1940
1950
1960
1970
1990
Log Population
(1)
Fraction Rural
(2)
Fraction on Farm
(3)
-0.081"*
(0.021)
-0.108**
(0.041)
-0.148*
(0.064)
-0.157
(0.080)
-0.127
(0.109)
-0.0118
(0.0090)
-0.0115
(0.0157)
-0.0165
(0.0199)
0.0084
(0.0225)
0.0159
(0.0095)
0.0204*
(0.0095)
-0.065**
(0.020)
-0.108**
(0.035)
-0.170**
(0.055)
-0.180*
(0.072)
-0.160
(0.103)
-0.0102
(0.0075)
-0.0065
(0.0126)
-0.0144
(0.0157)
-0.0094
(0.0181)
0.0000
(0.0049)
0.0077
(0.0070)
0.0334**
(0.0121)
0.0120
(0.0081)
-0.0027
(0.0096)
0.0004
(0.0065)
R-squared
0.5954
0.4558
0.9451
769
769
769
Sample Counties
Notes: For each indicated outcome variable, the column reports estimates from equation (2) in the text and described in the notes to
Table 2. All regressions are weighted by county population in 1930. Reported in parentheses are robust standard errors clustered by
county. ** denotes statistical significance at the 1% level, * at the 5% level.
Table 8. Changes After 1930 in Unemployment and Manufacturing
1930 Mean:
Erosion Level:
High - Low
1940
Unemployment
Rate
(1)
0.0410
[0.0255]
0.0071 *
(0.0034)
Workers perlabor force
(2)
0.0628
[0.0629]
0.0072
(0.0041)
Manufacturing
Workers
Log
per-capita
Establishments
(4)
(3)
0.0262
[0.0276]
0.0040
(0.0021)
1945
1950
1958 (-1964)
0.0036
(0.0061)
1967 (-1974)
0.0024
(0.0093)
1978 (-1982)
0.0050
(0.0109)
1987 (-1992)
-0.0016
(0.0025)
0.0000
(0.0035)
0.0005
(0.0020)
1945
1950
1967 (-1974)
0.0063
(0.0047)
0.0131
(0.0076)
1978 (-1982)
1987 (-1992)
0.108
-0.137
(0.078)
-0.126
(0.080)
-0.014
(0.099)
-0.036
(0.110)
-0.044
(0.129)
-0.138
(0.247)
-0.058
(0.261)
-0.226
(0.259)
-0.443
(0.494)
0.157
(0.362)
-0.080
(0.049)
-0.011
(0.047)
-0.127
(0.111)
0.003
(0.128)
-0.037
(0.058)
-0.100
(0.062)
-0.060
(0.079)
-0.061
(0.093)
-0.045
(0.109)
-0.041
(0.182)
-0.111
(0.178)
-0.153
(0.187)
0.022
(0.338)
-0.087
(0.263)
(0.145)
0.003
(0.183)
-0.0007
(0.0017)
1954
1958 (-1964)
0.022
(0.069)
-0.078
(0.066)
-0.0031
(0.0020)
1954
Medium - Low
1940
Log
Value Added
(5)
0.0250**
(0.0075)
R-squared
0.8550
0.5803
0.9203
0.5754
0.9456
Sample Counties
769
550
336
551
275
Notes: For each outcome variable, the 1930 mean is reported at the top of the column and the standard deviation in brackets. The
column reports estimates from equation (2) in the text and described in the notes to Table 2. In column 3, manufacturing worker data is
only available for 1958, 1967, and 1987, and population data is taken from the nearest decennial. Reported in parentheses are robust
standard errors clustered by county. ** denotes statistical significance at the 1%level, * at the 5% level.
Table 9. Log Changes After 1930 in Wage Proxies
Erosion Level:
High - Low
1940
Retail Sales
per-capita
(1)
Retail
Sector
(2)
Payroll divided by Workers
Manufacturing
Wholesale
Sector
Sector
(3)
(4)
-0.013
(0.018)
-0.035
(0.048)
-0.050
(0.067)
-0.049
(0.062)
-0.044
(0.051)
-0.050
(0.032)
-0.048
-0.059
-0.060**
(0.022)
-0.030*
(0.015)
-0.041 *
(0.016)
-0.085*
(0.038)
-0.035
(0.053)
(0.018)
(0.039)
-0.095
(0.069)
-0.130*
(0.060)
-0.209*
(0.104)
-0.043
(0.110)
-0.077
(0.065)
-0.063**
-0.012
(0.014)
-0.048
(0.034)
-0.131**
(0.049)
-0.098**
(0.029)
1945
1954
1958 (-1964)
-0.058
(0.031)
1967 (-1974)
-0.034
(0.039)
1978 (-1982)
1987 (-1992)
Medium - Low
1940
(0.022)
(0.035)
-0.202**
(0.067)
-0.104**
-0.116**
1945
(0.040)
-0.041 **
1954
(0.013)
1958 (-1964)
0.005
(0.021)
-0.017
(0.011)
1967 (-1974)
0.033
(0.028)
-0.016
(0.011)
0.011
(0.042)
-0.041
(0.024)
-0.014
(0.013)
1978 (-1982)
1987 (-1992)
R-squared
Sample Counties
0.9749
758
0.9970
748
-0.096**
(0.032)
-0.079**
(0.026)
-0.076**
(0.024)
-0.178**
(0.040)
-0.126**
(0.027)
0.9921
495
-0.092*
(0.045)
-0.106*
(0.044)
-0.073
(0.046)
-0.080
(0.072)
-0.065
(0.045)
0.9906
257
Notes: For each outcome variable, the column reports estimates from equation (2) in the text and described in the notes to Table 2. In
column 1, retail sales data is only available for 1958, 1967, and 1987, and population data is taken from the nearest decennial. Reported
in parentheses are robust standard errors clustered by county. ** denotes statistical significance at the 1% level, * at the 5% level.
Good Fences Make Good Neighbors:
Evidence on the Effects of Property Rights
Richard A. Hornbeck*
Abstract
The second essay examines the impact on agricultural development from the introduction of barbed wire fencing to the American Plains in the late
1 9 th
century.
Farmers were required to construct fences to be entitled to compensation for damage
by others' livestock. From 1880 to 1900, the introduction and universal adoption of
barbed wire greatly reduced the cost of fences, relative to predominant wooden fences,
most in counties with the least woodland. Over that period, counties with the least
woodland experienced substantial relative increases in settlement, land improvement,
land values, and the productivity and production share of crops most in need of protection. This increase in agricultural development appears partly to reflect farmers'
increased ability to protect their land from encroachment. States' inability to protect
this full bundle of property rights on the frontier, beyond providing formal land titles,
might have otherwise restricted agricultural development.
*I thank Michael Greenstone, as well as Daron Acemoglu, Lee Alston, Josh Angrist, David Autor, Abhijit
Banerjee, Dora Costa, Esther Duflo, Claudia Goldin, Tal Gross, Raymond Guiteras, Jeanne Lafortune, Philip
Oreopoulos, Paul Rhode, Peter Temin, and numerous seminar participants for their comments. I also thank
Daniel Sheehan for assistance with GIS software. hornbeck@mit.edu
The assignment and enforcement of property rights are thought to encourage economic efficiency [Coase 1960] and to be an important determinant of product choices, investment
levels, production methods [Cheung 1970], and macroeconomic performance [North 1981;
Acemoglu and Johnson 2005]. Coase presents the example of a farmer and a cattle-raiser,
in which the costless assignment and tradeability of property rights leads to the efficient
allocation of resources. Much attention has focused on the role of land rights in agricultural
development, where more secure rights might improve economic efficiency by increasing farmers' investments. 1 The typical positive correlation between investment and property rights,
however, may reflect the endogenous determination of property rights [Demsetz 1967; Miceli
et al. 2001].2
Valid causal estimation is dependent upon plausibly exogenous variation in property
rights, mainly either from instrumental variables [Besley 1995; Brasselle et al. 2002] or land
rights policies [Jacoby et al. 2002; Banerjee et al. 2002]. These exogeneity assumptions are
typically most plausible for micro-level variation in property rights, but this variation may
overstate the effects of larger reforms or macro-level differences.3
This paper attempts to estimate the effects of property rights on farmers' production
decisions, using a widespread and plausibly exogenous decrease in the cost of establishing
and enforcing property rights: the introduction of barbed wire fencing to the American Plains
in the late
1 9 th
century. At that time, farmers were required to build fences to protect their
land and crops from others' livestock. Without fences, farmers were apt to suffer substantial
damages and typically had no formal right to compensation from the livestock's owner.
Fences may also have decreased farmers' susceptibility to land expropriation by providing an
'See Besley [1995] for a discussion of three mechanisms: decreased expropriation raises the expected
return on investment; an improved ability to collateralize land increases access to credit; and lower costs of
trading land raise the expected return on investment.
2
Unobserved factors, such as land quality, may influence both investment and property rights. Investment
may increase the legitimacy of land claims or otherwise make the land more worthwhile to secure. Goldstein
and Udry [2005] discuss an opposite example where fallowing land may jeopardize ownership claims.
3
When property rights vary across land plots within small areas, observed effects may reflect sorting of
producers or substitution in production that would not occur if property rights were made more widely
available.
informal mechanism for delineating and substantiating land claims. Both effects of fencing
are tantamount to an increase in farmers' property rights in the standard security model of
property rights [Besley 1995]. 4
Before barbed wire, fence construction on the Plains had been severely restricted by high
costs in areas that lacked locally available fencing materials. Due to high transportation
costs from Eastern wooded areas, small sections of local woodland provided a vital source
of timber for fencing on the Plains. The introduction and universal adoption of barbed wire
from 1880 to 1900 had its greatest effect in areas with the least woodland that had been
most costly to fence. 5
Using decennial data from the Census of Agriculture, I estimate that counties with the
least woodland experienced large increases in agricultural development from 1880 to 1900,
relative to counties with sufficient woodland for farmers to have accommodated previous
fencing material shortages. Estimated increases were particularly substantial along intensive
margins of agricultural production. Controlling for constant differences among counties and
state-wide shocks to all counties, the fraction of county farmland that was improved increased
by 19 percentage points in counties with the least woodland.
From 1880 to 1890, average crop productivity increased by 23% in counties with the
least woodland, controlling for crop-specific differences among counties and crop-specific
state-wide shocks to all counties. The increased productivity was entirely in crops most
susceptible to damage from roaming livestock, as opposed to hay. In addition, farmers
shifted cropland allocations toward these more at-risk crops. Crop production changed little
from 1890 to 1900, indicating that farmers adjusted more rapidly along these margins.
Agricultural development increased along these intensive margins, even as counties with
the least woodland expanded along the extensive margin of total farmland. The estimates
4In contrast to other sources of variation in property rights, fencing did not affect farmers' formal land
ownership and, thus, had little effect on farmers' ability to trade the land or use it as collateral.
5
Barbed wire fencing need not affect property rights in other contexts with different legal structures or
geographical characteristics; rather, these factors combined to make barbed wire particularly important for
property rights protection in the United States' Plains in the late
and Hill [1975].
19 th
century. See, for example, Anderson
are robust to controlling for changes over time that are correlated with counties' distance
West and distance from St. Louis; counties' region, subregion, or soil group; or counties'
initial intensity of land improvement. By contrast, estimated increases along the extensive
margin of total farmland are more sensitive to these robustness checks.
Improvements in production were reflected in increased land values in the six Plains
states analyzed, totaling roughly 0.9% of national GDP. In all, these estimates lend support
to historical accounts that "without barbed wire the Plains homestead could never have
been protected from the grazing herds and therefore could not have been possible as an
agricultural unit" [Webb 1931, p.317].
The historical setting provides a rare opportunity to compare the effects of the de facto
improvement in property rights enforcement provided by barbed wire with the effects of
contemporaneous de jure efforts by some state legislatures. In the early 1870's, some states
and counties adopted "herd laws" that shifted formal fencing responsibilities to livestock
owners. In contrast to the large estimated effects of barbed wire, these reforms appear to
have had little effect. In this historical context when settlement was sparse, it appears that
the herd laws were weakly enforced and farmers continued to require fencing for protection.
Attributing the estimated effects of barbed wire to the increased security of property
rights requires that barbed wire affected agricultural development only by increasing property
rights enforcement. I examine the potential for two other effects of cheaper fencing: the direct
encouragement of cattle production and/or the joint production of crops and cattle. There
is little evidence of a positive effect of barbed wire on cattle production, however, and joint
production is found to have decreased. This fails to contradict the hypothesis that barbed
wire's effects operated mainly through an increase in property rights enforcement.
The paper is organized as follows. Section I reviews historical accounts of the need for
alternative fencing materials in timber-scarce areas and the introduction of barbed wire
fencing. Section II nests these historical accounts in a more stylized theoretical framework.
Section III describes the data and presents summary statistics. Section IV develops the
empirical methodology. Section V presents the main estimation results and explores their
robustness. Section VI examines other potential confounding factors. Section VII analyzes
states' reforms to fencing requirements. Section VIII concludes.
I
History of Barbed Wire and the Great Plains
I.A
Timber Shortages Constrained Property Rights Enforcement
Breaking with English common law, legal codes in the United States in the
1 9 th
century had
long required the construction of a "lawful fence" 6 to secure farmland [Kawashima 1994].
Without such a fence, a farmer had no formal entitlement to compensation for damages
inflicted by roaming livestock. In practice, fences were necessary to protect crops and they
required substantial investment. In 1872, fencing capital stocks in the United States were
roughly equal to the value of all livestock, the national debt [U.S. House 1872], or the
railroads. Annual fencing repair costs were greater than the combined annual tax receipts
at all levels of government [Galveston News 1873, as cited in Webb, pp.288-289].
Fencing became increasingly costly as settlement moved into areas with little woodland.
As early as 1841, agricultural journals discussed the lack of timber for fencing as the major
barrier to the settlement of Western regions [Bogue 1963a, p.74]. An 1871 report, prepared as
a guide for immigrants, focused on three main characteristics of farmland in Plains counties:
its price, the amount of timber, and the amount fenced [U.S. House 1871].
When the frontier line left the timbered region and came onto the Prairie Plains,
the pioneers found there neither timber nor stone. There was nothing with which
to fence the land. ... Without fences [they] could have no crops; yet the expense
of fencing was prohibitive, especially in the Plains proper. [Webb, p.281, p.287]
When he sought to fence his crops against marauding livestock, the prairie farmer
faced the timber problem at its most acute. [Bogue 1963b, p.7]
6
The exact requirements for a lawful fence varied by state, but a fence was generally required to be strong
enough to prevent the intrusion of a roaming animal. Some laws also required that the animal's owner be
shown to have contributed to the damages through particular negligence or maliciousness.
During the early period of settlement there was little conflict between those growing crops
and those raising cattle, but that was because farmers were mostly forced out of necessity to
remain elsewhere. Much potential conflict existed [Alston et al. 1998], which is the relevant
factor in farmers' location and production decisions. 7
High transportation costs made it impractical to supply areas without woodland with the
required amounts of timber for fencing [Kraenzel 1955, p.129; Hayter 1939; Bogue 1963b,
p.7], though systematic data on transportation costs are unavailable. In low-woodland areas,
"the scarcity of wood and its consequent high cost encouraged experimentation" [Primack
1969, p.287]. There was some use of hedge fences, yet these were found to be too difficult
and costly to grow and control. Smooth iron wire fences were available, yet they were easily
broken by animals, prone to rust, and broke (sagged) in cold (hot) weather.
Writing about central Iowa, Bogue [1963b, p.6] further describes which counties were
most affected by timber shortages:
Where timber and prairie alternated, locations in or near wooded areas were
relatively much more attractive.... [T]here developed a landholding pattern of
which the timber lot was an intricate part. Settlers on the prairie purchased five
or ten acres along the stream bottoms or in the prairie groves and drove five, ten
or fifteen miles to cut building timber or to split rails during the winter months.
Smaller counties were roughly 30 miles on each side, so farmers traveling 5-15 miles for timber
would still have been mostly within their home county. For simplicity and transparency, I
assume that farmers' access to wooden fence materials depended on the amount of woodland
in their home county.
The empirical analysis will examine continuous variation in county woodland levels, but
Bogue's description provides some useful benchmark values for the level of woodland in
a county. For a standard homestead farm size of 160 acres, a county would need to be
7
Differences in the ability to prevent encroachment are key, rather than differences in the observed initial
equilibrium levels of encroachment. The empirical analysis does not define counties with a "property rights
problem" as those with both livestock and cropland, as this is endogenously determined by many unobserved
local factors. Instead, the paper focuses on local woodland as an exogenous determinant of fencing availability
and the cost of securing farmland.
roughly 4% woodland for each farm to acquire 5-10 acres of woodland. Based on this
calculation, counties are assigned to the following three woodland categories: low (0-4%),
medium (4-8%), and high (8-12%). The designated "low" counties roughly represent those
most constrained by timber scarcities, while "medium" counties could have partially adjusted
with this landholding pattern and then have been less affected along with "high" counties.
The exact cutoffs for these categories are not relevant for the results; rather the continuous
estimates will be evaluated at three corresponding benchmark levels (0%, 6%, 12%) in order
to assist in interpreting the estimated magnitudes.
I.B
The Arrival of Cheap Barbed Wire Fences, 1880 to 1900
Crude versions of enhanced wires were patented as early as 1867, but the most practical and
ultimately successful design for barbed wire was patented in 1874 by Glidden, a farmer in
DeKalb, Illinois. Glidden's design, which featured two twisted strands of wire that held metal
barbs in place, had three important characteristics: barbs prevented cattle from breaking the
fence; twisted wires tolerated temperature changes; and the design was easy to manufacture.
Glidden sold a half-stake in the patent for only a few hundred dollars to Ellwood, a hardware
merchant in DeKalb, and the two started the first commercial production of barbed wire,
producing a few thousand pounds per year by hand [Hayter].
Barbed wire was cheaper than wooden fencing, particularly in timber-scarce areas, and it
had lower labor requirements.8 Correspondence from Glidden and Ellwood in 1875 reflected
the close relationship between local woodland and the demand for barbed wire: "where
lumber is reportedly dearer, the wire would probably sell for more" [Webb p.310]. Given
current prices, they did "not expect the wire to be much in demand where farmers can build
brush and pole fences out of the growth on their own land" [Hayter].
sIn Iowa, wooden fences varied in total construction costs per rod from $0.91-$1.31 in 1871, while barbed
wire fences cost $0.60 in 1874 and below $0.30 in 1885 [Bogue 1963b, p.8]. Other reports quote barbed wire
fences as costing $0.75 per rod in Indiana in 1880, while hedge fences cost $0.90 per rod and were wasteful
of the land [Primack 1977, p.73]. Primack [1977, p.82] estimates that a rod of barbed wire took 0.08, 0.06,
and 0.04 days to construct in 1880, 1900, and 1910. The labor requirements for constructing wooden fences
were constant throughout this period: 0.20, 0.34, and 0.40 days for board, post and rail, and Virginia rail.
In 1876, the country's largest plain wire manufacturer (Washburn & Moen) bought half
of Glidden's business for $60,000 cash, plus royalties, and began the first large-scale production of barbed wire. By 1879, they had produced nine million pounds. 9 In contrast to
Glidden's sale to Ellwood, Washburn & Moen's purchase showed an awareness of barbed
wire's potential. Even the initial level of demand was sufficient to ensure "enormous profits"
[Webb, p.309]. 10
Newspaper advertisements began to appear in Kansas and Nebraska in 1878 and 1879
[Davis 1973, pp.133-134] and there was a series of public demonstrations in the early 1880s.
Once the effectiveness of barbed wire was proved, "Glidden himself could hardly realize
the magnitude of his business. One day he received an order for a hundred tons; 'he was
dumbfounded and telegraphed to the purchaser asking if his order should not read one
hundred pounds' " [Webb, p.312].
Table 1 provides statistics on the timing of barbed wire's introduction and its universal
adoption. Panel A shows a sharp increase around 1880 in the annual production of barbed
wire. Panel B shows the resulting transformation in regional fence stocks. Before 1880, fences
were predominately made of wood. From 1870 to 1880, there were some small increases in
wire fencing, including both smooth wire and barbed wire. After 1880, there were rapid
increases in barbed wire fencing. Total fencing increased most in the Plains and Southwest
regions where there were more timber-scarce areas. Wood fencing also initially increased,
however, highlighting that it would be inappropriate to attribute all regional increases in
fencing and economic activity to the introduction of barbed wire.
Even as the quality of barbed wire improved and consumers became increasingly aware
of its effectiveness in the early 1880s, falling input costs and manufacturing improvements
9
This process began in 1875 when Washburn & Moen, headquartered in Massachusetts, sent an agent to
investigate unusually large orders from DeKalb, Illinois. They acquired barbed wire samples and contracted
an expert machinist to design automatic machines for its production.
10
McFadden [1978] provides details on the development of these businesses, with the 1899 incorporation
of the American Steel and Wire Company of New Jersey leading to the monopolization of the barbed wire
industry.
drove down prices: $20 (1874), $10 (1880), $4.20 (1885), $3.45 (1890), and $1.80 (1897).11
Panel C of Table 1 shows that all new fences constructed after 1900 were made of barbed
wire, so further price declines or quality improvements would not have had a differential
effect across counties with varying access to wooden fences. Thus, the differential effects of
barbed wire on farmers' fencing costs were predominately from 1880 to 1900.12
The empirical approach presented here requires that the introduction of barbed wire
fencing was exogenous, i.e., that its rapid rise around 1880 was not caused by the anticipated
development of low-woodland areas. This assumption appears plausible for two main reasons.
First, from a microeconomic perspective, the demand for fencing alternatives had already
been high for decades and Glidden-Ellwood appear not to have anticipated the tremendous
market demand for barbed wire.
Second, from a more macroeconomic perspective, the
necessary cheap steel was only becoming available around 1880.
Barbed wire's widespread commercial success was made possible by unrelated developments in the industrial steel sector. The production of strong rust-free barbed wire required
steel, which became dramatically cheaper as the Bessemer-steel process became widely used
(patented in England in 1855).
Figure 1 shows prices for iron, steel, and barbed wire.
Barbed wire's introduction and mass-production follows the sharp decline in steel prices in
the 1870s. 13 Primack [1969] summarizes:
Outcries about the burdens of fencing by agriculturists in the 1850 to 1880 period
seem amply justified. A need was revealed and the problem was resolved, not by
changing laws and institutions but rather by technological change. This solution
had to wait for the development of cheap steel in the industrial sector. Then a
solution was found in wire fencing, cheap in both money and labor costs. [p.289]
11
Prices
2
are per hundred pounds [Webb p.310]. Hayter reports similar prices for 1874 and 1893.
1 Even by 1890, new fence construction in much of the country was entirely of barbed wire, yet there was
still some wooden fence construction in the Prairie and the Southwest. This likely reflects less-developed
distribution networks, as well as ranchers' opposition to enclosure and fence-cutting wars that were resolved
in the late 1880s [McCallum and McCallum 1965, pp.159-166; Webb, pp.312-316].
13
Previously discussed historical accounts provided additional price observations for barbed wire, indicated
with X's, though pre-mass production prices are off the chart. Note that the 1899 increase in barbed wire
prices coincides with the monopolization of the barbed wire industry, though the Spanish-American War
may have contributed to the observed increases in prices for barbed wire, steel, and iron.
II
Theoretical Framework
To provide these historical accounts a more formal theoretical structure for clarity and estimation purposes, consider a farmer in each county c and time period t choosing a level of
investment Id and property rights enforcement Rct to maximize profits. The farmer produces output F(It,qct), where qct denotes the quality of the land or a composite of other
characteristics.
F(., -) is increasing in both arguments and F 1 2 (', ) > 0. Following Besley
[1995], an expected percentage of output is lost in each period, T(Ret) E [0, 1], and it is
decreasing in the level of property rights enforcement, T'(Rct) < 0. Investment and property
rights are each produced at some cost, Cct(I t) and Cct(Ret). Thus, farmers choose Ict and
Rt to maximize:
(1)
(1 - r(Rct))F(Ict, qct) - C(Ict) - Oct(Rct).
An optimal interior solution satisfies two first-order constraints:
(2)
(3)
C't(Ict) =
Ct(Rt)
=
Fl(Ict,qct)(1 - T(Rt))
-r'(Rt)F(Ict, qt).
In equation (2), the marginal cost of investment is set equal to the amount of marginal return
that the farmer expects to retain. In equation (3), the marginal cost of property rights is
set equal to the marginal increase in retained total output. These equations generate three
relationships of interest. First, the optimal choice of investment is increasing in the level
of property rights because a greater proportion of the returns would be kept. Second, the
optimal choice of property rights is increasing in the level of investment because there is
an added incentive to keep the output. Third, higher land quality would directly increase
both property rights and investment by raising total production and the marginal return to
investment. The empirical identification problem is that an observed correlation between Ict
and Rct could reflect more than one of these three effects.
It may be possible, however, to identify the direct effect of property rights on investment
by examining the effect of changes in the marginal cost of producing property rights. Equation (2) defines the optimal choice of investment, It(R*, qt), and inserting that function into
equation (3) defines R,*(qct, Ct(R* ) . Given some equilibrium marginal cost Pet of establishing property rights, it follows that dl*= I*
.R* Thus, an estimate of
would identify
up to some negative term, R*. Empirical analyses rarely have a natural or comparable
unit of measurement for property rights, in which case a "reduced form" estimate of dI* is
as informative about the underlying causal relationship
f
as if divided by a "first-stage"
estimate of oR*
aP
To relate the marginal cost of producing property rights to the introduction of barbed
wire, assume further that property rights are produced through the use of some combination
of timber and barbed wire, Rct = R(Tt, Bet). The price of barbed wire (pB ) is assumed to
be decreasing over time, but constant across counties. 14 The price of timber in each county
(pT) is assumed to be constant over time, but decreasing in the percentage of the county
that is woodland, i.e., p = g(W) and g'(We) < 0.15 The cost of producing property rights
is the minimum obtainable value from choosing B
(4)
and Tt to minimize
pB . B + p .-Tt, subject to: Ret(Ta, Bt) = R.
If timber fences are initially constructed and they are not a perfect complement to barbed
wire, a decrease in the price of barbed wire that results in its construction would decrease
the marginal cost of producing property rights more in counties with less woodland and
higher timber prices, i.e.,
pTOpa
> 0. Once timber fences were no longer constructed,
further price declines would have no differential effect across counties with different woodland
14
While demand for barbed wire was higher in timber-scarce regions, the market for barbed wire became
sufficiently competitive and spread throughout the country that differences in local prices would have been
minimal.
15
Local timber prices may have changed over time, at least in part due to barbed wire, but the subsequent
universal adoption of barbed wire implies that changes in timber prices had little differential effect on fencing
costs.
levels."
Thus, barbed wire would encourage higher levels of property rights protection in
timber-scarce areas in the period from its widespread introduction to its universal adoption
(1880-1900).
III
III.A
Data and Summary Statistics
Data Construction
The dataset is drawn from the United States Census of Agriculture and includes county-level
data by decade from 1870 to 1920 [Gutmann 2005]. The analysis is restricted to Plains states
for which data are available in every census from 1870 to 1920: counties in Minnesota, Iowa,
Nebraska, Kansas, Texas, and Colorado. Data in these counties are first available in 1870
and data on land improvement are available through 1920. Some county boundaries changed
over this period, and the data are adjusted to hold the 1870 geographical units constant (see
Data Appendix).1 7
Figure 2 shows the sample counties, shaded to represent initial woodland levels that are
held fixed in the empirical analysis. l s Because the empirical analysis will include stateby-decade fixed effects, the relevant variation is mostly in Iowa, southern Minnesota, and
the eastern parts of Kansas and Nebraska. Non-sample counties in these states are mainly
excluded because of unavailable data in 1870 and 1880.
16This case represents a corner solution to equation (4).
17
Data starting in 1870 are not available for other Plains states, such as Oklahoma, South Dakota, and
North Dakota. Partial data from states further East are available from other sources, but local variation in
woodland had less influence on the availability of fencing materials and the remaining woodland would have
been more likely to reflect differences in previous land development.
18
Data are available for the number of acres of woodland in farms, so county woodland levels are defined
as initial acres of woodland in farms per county acre. This is not ideal because low woodland may reflect low
initial settlement. For this reason, initial woodland levels are calculated using 1880 data when settlement
was higher. Because wooded areas in these counties were initially most desirable, it seems plausible that
most county woodland would be reflected in this measure. The empirical results are similar when using 1870
woodland levels. Later woodland levels are potentially endogenous to changes in agricultural development,
but these results are also similar.
The empirical analysis focuses initially on two measures of farmers' land-use: the percent
of county land that is in farms, including land for crops and livestock; 19 and the percent
of land in farms that is improved. 20
The percent of county land in farms represents the
extensive margin, which mainly reflects farmers' expected returns to converting land from
the public domain.
The percent of farmland improved represents the intensive margin,
which may reflect farmers' willingness to sink investments into the future productivity of
that land. Improved land could be plowed for crops or otherwise prepared for livestock, but
the definition appears to exclude land that is simply fenced.
In focusing on the improvement intensity of farmland, this paper is similar to most farmlevel studies of property rights that analyze whether a given plot of farmland has been
improved in some way. An increase in this measure may reflect that farmers substituted
toward improvement-intensive activities, where profitability is typically more sensitive to
expropriation risk.21
The data are available by decade, so there is limited flexibility in analyzing responses
to the exact timing of barbed wire's introduction. Because the mass distribution of barbed
wire was just beginning by 1880 and fencing stocks had yet to respond substantially, all
1880 county outcomes represent the end of the pre-barbed wire period.22 Because new fence
construction was entirely barbed wire by 1900, at the latest, 1900 marks the moment when
barbed wire no longer had a differential effect.
19
The definition of land in farms differs very slightly in some periods, but generally "describes the number
of acres of land devoted to considerable nurseries, orchards and market-gardens, which are owned by separate
parties, which are cultivated for pecuniary profit, and employ as much as the labor of one able-bodied workman during the year. To be included are wood-lots, sheep-pastures, and cleared land used for grazing, grass
or tillage, or lying fallow. Those lands not included in this variable are cabbage and potato patches, family
vegetable-gardens, ornamental lawns, irreclaimable marshes, and considerable bodies of water" [Gutmann
2005].
20
The clearest definition of improved land is from the 1920 Census: "All land regularly tilled or mowed,
land in pasture which has been cleared or tilled, land lying fallow, land in nurseries, gardens, vineyards, and
orchards, and land occupied by farm buildings" [Gutmann 2005].
21
If it were true that investment increased property rights but property rights did not encourage investment
[Brasselle et al.], then the intensity of land improvement should decrease when barbed wire offered a cheaper
alternative for protecting property rights.
22
Land-use measures were reported for the Census year and productivity is imputed from production and
acreage in the previous year.
III.B
Summary Statistics
The average amount of woodland in the sample is 10%, but most counties have lower woodland levels: 39% have 0-4% woodland, 15% have 4-8% woodland, 11% have 8-12% woodland, and 35% have more than 12% woodland. At the lower, middle, and upper end of the
first three categories, three benchmarks (0%, 6%, 12%) represent meaningful points in the
distribution of woodland levels: 20% of the sample has less than 1% woodland; the median
woodland level is 6%; and 12% is among the higher typical levels of woodland.
Over the entire sample region, counties with different woodland levels are not evenly
balanced geographically. Thus, the main results all focus on specifications that include a
full set of state-by-decade fixed effects. To account for geographic differences within states,
I present results from specifications that also control for distance West, distance from St.
Louis, or finer regional groupings.
Table 2 reports average county characteristics in 1880 for all sample counties and within
the first three woodland categories. Prior to barbed wire's introduction, low-woodland counties tended to be larger but lagged behind in settlement, land value, and agricultural production. Less of the farmland in low-woodland counties was improved or used for crops.
Low-woodland counties had somewhat less cropland in corn and more in hay, which is less
susceptible to damage from livestock. Overall, crop choices are fairly balanced across the
three categories, however, which somewhat mitigates concerns that crop-specific technology
or price shocks could have large differential effects by woodland percentage.
Total fencing expenditures were lower in low-woodland counties, but were roughly 4% of
the total output value across each woodland percentage group. Given higher per unit costs in
low-woodland areas, this would suggest a lower intensity of fencing in those areas. 23 Mediumwoodland county averages generally fall between those for low-woodland and high-woodland
counties, but are much closer to the latter.
23
Data on fencing expenditures was only collected in 1880.
IV
IV.A
Measurement Framework
Estimation Setup: A Discrete Example
The estimation strategy is most easily described in a discrete example with two county
types and two time periods. Assume that farmers in low-woodland counties with W, E
(0, 0.04) have a high timber price pT, while farmers in medium-woodland counties with
W E (0.04, 0.08) have a medium timber price pT.
Furthermore, the price of barbed wire
is infinite in the first time period and is pB in the second time period, with pB < ph. A
potential estimator for the effect on the production outcome, Y, from a decrease in the cost
of property rights protection is then:
(5)
(Yc=L,t=2 -
c=L,t=) -
(c=M,t=2
-
c=M,t=1)
For this difference-in-difference estimator to be unbiased, farmers' production decisions
must depend additively on unobserved factors, such that:
(6)
Y*t(R*t, qtt)= F(R
1 ) + 7t + pI + cct
and E[Ec] = 0. The assumption is that farmers in counties with different amounts of woodland would have, on average, made the same production changes if not for an increase in
property rights protection due to barbed wire's introduction. The analysis in section II was
substantially simplified by excluding some variables from each function, such as in separating the costs of property rights and investment, but any potential for estimation bias can be
thought of as a violation of this assumption.
It is impossible to test this identification assumption directly, but additional time periods
and greater variation in county types can be used to form indirect tests. First, the same
estimator for two periods before the introduction of barbed wire tests whether investment
decisions in these two types of counties had been trending similarly. Second, the same
estimator for two periods after the universal adoption of barbed wire would test for other
sources of differences, given that further price declines would not have differential effects.
Any differential trends before barbed wire's introduction or after its universal adoption may
or may not have occurred between those periods, but the results can be tested for robustness
to each scenario.
A third specification test is based on the presumption that local wooden fencing costs
depended on local woodland in a non-linear and convex manner. Once settlers were able
to acquire the needed acres of woodland, as described by Bogue, one would expect further
woodland to have made less difference in the cost of fencing materials.2 4 If high-woodland
counties with W, E (0.08,0.12) had a low timber price pT that was much closer to pT
than was pH, then the estimate from equation (5) should be greater than a similar estimate
comparing medium- and high-woodland counties. This test is performed by the differencein-difference-in-difference estimator:
(7)
[(Ic=L,t=2 -
c=L,t=1)
-
(Ic=M,t=2 -
c=M,t=1)]
-
[(Ic=M,t=2 -
Ic=M,t=l) -
(Ic=H,t=2 -
Ic=H,t=l)].
The intuition for this analysis can be seen in a plot of the average improvement intensity
of farmland, broken out by decade and county woodland group. In Figure 3, medium- and
high-woodland counties are shown to have changed similarly over the entire period. Lowwoodland counties also changed similarly, except for large relative increases during the period
of adjustment to barbed wire (1880-1900). This analysis of woodland categories is intended
only to illustrate the intuition for the methodology, whereas the later empirical analysis
examines continuous variation in woodland levels and provides a more thorough treatment
of potential confounding factors.
24
In particular, as more wooded areas become scattered throughout a county, the distance to the nearest
plot would tend to fall at a decreasing rate. Also, it was increasingly difficult to adjust the type of wooden
fence to substitute away from using as much timber [Primack 1977, p.70].
IV.B
Main Estimating Equation
For the main empirical analysis, county-level production outcomes are first-differenced to
control for any county characteristics that are constant over time.2 5 State-by-decade fixed
effects a st are included to control for state-specific shocks that have an equal effect on all
counties in the state. To allow flexibly for changes over time to be correlated with county
woodland levels, included for each decade is a fourth-degree polynomial function of a county's
woodland level in 1880. For each production outcome, the estimated equation is:
T
(8)
Yt - Yc(t-1) = Cost + E(3itWc +
02tW2c
+
3 tWc3 + 34tW
4
c)
Ect
t=2
The estimated p's summarize how changes over each decade in production outcome Y varied
by county woodland level W.
V
V.A
Estimation Results
Land Improvement and Land Settlement
Equation (8) is estimated for the fraction of farmland improved in each county. The full set
of estimated O's is prohibitively difficult to interpret numerically, but the results can be seen
in Figure 4.26 The solid line reports the estimated change over the indicated time period for
25
Estimating equation (8) with county fixed effects instead yields identical estimates, but their standard
errors tend to be higher by 8-10% because the untransformed error terms are closer to being a random walk
than serially uncorrelated. To see this, consider that if equation (8) were estimated for each time period
without first-differencing Y, changes over time in the estimated coefficients would still not depend on fixed
county characteristics. That is, whether some county with high woodland always had a high outcome level
would not influence estimated changes in the dependence of that outcome on woodland. Thus, adding county
fixed effects to the regression does not affect differences in the estimated coefficients across periods, though
it does increase their precision by reducing unexplained variation. Those exact differences in estimated
coefficients are obtained by first-differencing Y and, in this case, it further reduces the unexplained variation
in Y.
26
For conciseness, the displayed results are limited to woodland levels less than 0.12 or 12%, though
equation (8) is estimated for the entire distribution of woodland levels.
a county with that woodland level, relative to the estimated change for a county with 0%
woodland. 27 The two dashed lines report 95% confidence intervals around the estimates.
From 1880 to 1890 and 1890 to 1900, counties with the least woodland made large relative
gains in the improvement intensity of farmland. By contrast, there were no substantial
relative changes at low woodland levels before 1880, after 1900, or at higher woodland levels
from 1880 to 1900.
In order to display and interpret these results numerically, the estimated changes are
evaluated at representative woodland levels: the most affected low-woodland county, with
0% woodland; the average medium-woodland county, with 6% woodland; and the least
affected high-woodland county, with 12% woodland. The predicted change for a county with
0% woodland relative to the predicted change for a county with 6% woodland is analagous
to a difference-in-difference estimate for counties with those exact woodland levels, but the
parameterized regression uses available data from counties with similar woodland levels.
Columns 1 and 2 of Table 3 report the evaluated numerical results from estimating
equation (8) for the fraction of farmland improved. In each decade, the coefficient in column
1 corresponds exactly to the difference in the graphed solid line at 0 and 0.06 in Figure 4.
The estimated magnitude is interpreted as follows: the top coefficient in column 1 reports
that acres of improved land per acre of farmland increased from 1870 to 1880, on average, 1.5
percentage points more in a county with 0% woodland than in a county with 6% woodland. 28
From the same regression, column 2 reports the predicted change for a county with 6%
woodland relative to a county with 12% woodland.
In parentheses is the standard error
for each coefficient, corrected for heteroskedasticity and clustered at the county level. In
brackets is the t-statistic of the absolute difference between the coefficients comparing 0%
vs. 6% and 6% vs. 12%. For example, the coefficients in the first row of columns 1 and 2 are
not statistically different with a t-statistic of 0.22.
2 7Due
to the inclusion of state-decade fixed effects, the estimated results are only interpretable relative to
some defined benchmark woodland level.
2 8That is, a county with 0%woodland that had
50% of its farmland improved would have, in expectation,
caught up to a county with 6% woodland that initially had 51.5% of its farmland improved.
From 1880 to 1900, the improvement intensity of farmland increased by a statistically
significant and substantial 19 percentage points in counties with 0% woodland relative to
counties with 6% woodland (Table 3, column 1). Recall that from its widespread introduction
in 1880 to its universal adoption by 1900, barbed wire most increased the availability of
fencing in counties with the least woodland. By contrast, there are few substantial estimated
changes before 1880, after 1900, or between higher woodland levels from 1880 to 1900. This
result is clear in Figure 5, which plots the estimated cumulative changes after 1870.
It lends credibility to the results that there are few substantial changes apart from the
estimated increases at low woodland levels from 1880 to 1900. Depending on assumptions
about other agricultural changes, the reported relative changes from 1880 to 1900 could
themselves be differenced across and/or within time periods. If the relative trends from
1870 to 1880 would have continued, the adjusted increase from 1880 to 1900 would be 16
percentage points for a county with 0% woodland relative to a county with 6% woodland.2 9
If estimated changes at higher woodland levels (6% vs. 12%) reflect other contemporaneous
changes that were linearly correlated with woodland levels, the adjusted change from 1880
to 1900 would be 17 percentage points.3 0 Under both assumptions, the adjusted change from
1880 to 1900 would again be 19 percentage points.3 1
The increase in the improvement intensity of farmland came despite substantial amounts
of land being converted from the public domain to private farmland. Columns 3 and 4 of
Table 3 report the results from estimating equation (8) for the fraction of county land in
farms. In these baseline estimates, settlement increased by 26 percentage points from 1880
to 1900 in counties with 0% woodland relative to counties with 6% woodland. There were
29
This 16 percentage point change reflects the increase from 1880 to 1890, plus the increase from 1890 to
1900,
minus two times the increase from 1870 to 1880.
30
This 17 percentage point change reflects the increase from 1880 to 1900 for a county with 0% woodland
relative to a county with 6% woodland, minus the change from 1880 to 1900 for a county with 6% woodland
relative to a county with 12% woodland.
31
This 19 percentage point change reflects the increase from 1880 to 1900 for a county with 0% woodland
relative to a county with 6% woodland (minus two times the increase from 1870 to 1880), minus the change
from 1880 to 1900 for a county with 6% woodland relative to a county with 12% woodland (minus two times
the increase from 1870 to 1880).
also some relative increases from 1870 to 1880, however, and from 1880 to 1900 counties
with 6% woodland made substantial relative gains on counties with 12% woodland. These
estimated changes in settlement are quite large, perhaps implausibly so, and are indeed more
sensitive to robustness checks than are the estimated changes in improvement intensity.
Robustness Tests
Table 4 presents the results from specification tests that include control variables meant
to capture other changes in Western agricultural development. The results are condensed to
show only the estimated changes from 1880 to 1890 and from 1890 to 1900 in counties with
0% woodland relative to counties with 6% woodland. For completeness, the full results for
Table 4 are shown in Appendix Tables 1, 2, and 3. As a basis for comparison, column 1 of
Table 4 reports the corresponding estimates from Table 3.
(i) Westward Development.
One concern is that the baseline estimates could be
confounded with an independent push toward increased westward development. From Figure
2, counties with less woodland tend to be further West than counties with more woodland,
even within states.
To address this concern, equation (8) is estimated with a control variable for the distance
West of each county center, interacted with each decade.3 2 Column 2 of Table 4 reports the
primary results from this specification, while the full results are reported in columns 2a-2d
of Appendix Table 1.
For the improvement intensity of farmland (Panel A), the baseline results are quite robust
to controlling for Westward development in each time period. For the amount of farmland
(Panel B), the estimated magnitudes fall by roughly half. Indeed, there are substantial
Westward settlement effects in each decade from 1870 to 1900 (not shown in the tables). By
contrast, there is very little Westward change in the improvement intensity of farmland from
1880 to 1900, though some moderate insignificant increases from 1870 to 1880.
32
To be precise, "distance West" refers to the x-coordinate of the county centroid on an equal area map
projection of the United States.
The "Westward development" of the United States may not have been just a move West,
but also an expansion from the middle. To allow for this possibility, an additional control
variable is included for the distance of each county center to St. Louis (the "Gateway to the
West"), interacted with each decade. Column 3 of Table 4 reports these results, which are
similar to controlling only for distance West. Note that for both specifications, Appendix
Table 1 reports little substantive change in the improvement intensity of farmland before
1880, after 1900, or between higher woodland levels from 1880 to 1900.
Linear distance coefficients are easiest to interpret, but these robustness results are similar
when controlling for non-linear changes in Westward development (distance West squared,
distance West cubed, distance from St. Louis squared, distance from St. Louis cubed). 3 3 In
these specifications, Colorado and Texas are contributing to the estimation of the distance
controls even though these states contribute little to the estimation of the woodland variables.
These robustness results are also similar when excluding sample counties in Colorado and
Texas, so that the distance controls and woodland variables are estimated primarily on the
same sample of counties.
Overall, it does not seem that Westward development accounted for the estimated increases in land improvement from 1880 to 1900 in counties with the least woodland. By
contrast, Westward development may account for a substantial portion of the baseline estimated increase in total farmland.
(ii) Regional Development. Another concern is that low-woodland counties may be in
different resource regions or soil groups, where they could be affected differently by changes
in product prices and technologies [see, e.g., Olmstead and Rhode 2002]. To explore this
possibility, the county boundary files were merged with two maps indicating land resource
regions and subregions, and great soil groups.3 4 Separate variables are defined for the percent
33
Note that whenever I refer to results being similar, that refers both to the coefficients and their precision.
Land resource regions and subregions were mapped in the 1966 U.S. Department of Agriculture Handbook 296. Soil groups were mapped by the U.S. Soil Conservation Service in 1951 and were retained in
the National Archives Record Group 114, item 148. These maps were scanned, traced in GIS software, and
digitally merged to 1870 county boundaries.
34
of each county area falling into each region, subregion, or soil group. Within the sample of
counties, there are 11 regions, 43 subregions, and 19 soil groups. The soil groups appear also
to capture general differences in climate and other environmental factors; in particular, one
soil group effectively corresponds to the presence of a major river.
To allow for differential growth patterns across each region, equation (8) is estimated
with an additional quadratic time trend for each of the 11 regions.3 5 The primary results are
reported in column 4 of Table 4. Similarly, columns 5 and 6 of Table 4 report the primary
results when controlling for quadratic time trends for each of the 43 subregions and 19 soil
groups, respectively.
For the improvement intensity of farmland, the results are fairly robust to each of these
specifications. The estimated magnitudes decline most when controlling for quadratic time
trends for each of the 43 subregions, but by less than one standard error. The estimates
are substantively similar to the baseline, remain statistically significant, and are statistically
greater than the estimates at higher woodland levels. Appendix Table 2 reports the full
results, which are also similar. For the total amount of farmland, the estimated increases
are less robust but still substantial and statistically significant.
Instead of including quadratic time trends, it is possible to include a decade fixed effect
for each region, subregion, or soil group. Although demanding of the data, especially for the
43 subregions, the results are mostly consistent with the baseline results.3 6
In summary, it appears that the baseline results cannot be explained by broad differences
in the regional balance among counties with different levels of woodland. Estimated changes
in improvement intensity are robust to finer regional controls, while changes in total farmland
appear to be more sensitive.
35
To be precise, the first-differenced analog of a quadratic time trend is included.
When including region-by-decade controls, the results are similar. When including soil-by-decade controls, the changes from 1880 to 1900 are similar but there is also a substantial increase in improvement
intensity from 1870 to 1880. In contrast to the 1880 to 1900 increases, however, this 1870 to 1880 increase
is not statistically greater than the increase at higher woodland levels. When including subregion-by-decade
36
controls, improvement intensity has similar increases from 1890 to 1900 but only a small insignificant increase
from 1880 to 1890. There is a moderate increase from 1870 to 1880, but not statistically greater than at
higher woodland levels.
(iii) Convergence in Agricultural Development. Another concern is that the baseline estimates could be confounded with convergence in agricultural development, whereby
counties with lower initial levels of land improvement or settlement may have naturally experienced higher subsequent growth. Counties with little woodland initially lagged behind
counties with more woodland, though there were only small relative increases from 1870 to
1880 when these initial differences were greatest.
To address this concern, equation (8) is estimated with an additional fourth-degree polynomial function of the county's fixed 1870 outcome level, interacted with each decade. 37
This effectively restricts the analysis to counties with different woodland levels but similar
outcome levels in 1870. 38 The primary results are reported in column 7 of Table 4, while the
full results are in Appendix Table 3.
For the improvement intensity of farmland, the estimated increases from 1880 to 1900
are approximately three-fourths their original size, statistically significant, and statistically
greater than changes at higher woodland levels (Table 4, column 7, Panel A). These increases
reverse the substantial relative decline before barbed wire's introduction and smaller but
statistically significant declines after barbed wire's universal adoption (Appendix Table 3,
column 1).3 9 For the amount of total farmland, by contrast, the estimates are substantially
smaller and now statistically insignificant. These increases would only seem substantial if
the negative pre-trend from 1870 to 1880 were subtracted from the subsequent changes.
37
The estimates are similar for alternative polynomial functions of the initial outcome level, but it is
appealing to use the same function for both the initial outcome level and woodland level.
38This robustness check has some statistical appeal, but it is less motivated by economic theory. Two
counties with the same initial outcome level and different woodland levels might be expected to differ along
important dimensions, in order for farmers in the lower woodland county to be compensated for the lack of
woodland. For example, prior to barbed wire, farmers might only improve as much land in a county with little
woodland as in a county with more woodland if the county with little woodland had higher quality soil to
justify the higher costs of fencing. The ideal empirical analysis would control for all important differences in
county characteristics, but this partial statistical correction may be counterproductive in identifying counties
with different woodland levels that were otherwise similar.
39
For the improvement intensity of farmland, it is worth noting the results when controlling for the initial
outcome level and implementing the other specification changes discussed above. The results are generally
similar to those reported in Appendix Table 3, with the notable exception that the estimated increase from
1880 to 1890 varies between zero and its original baseline estimate.
Overall, these estimates are consistent with previous findings that counties with the least
woodland experienced relative increases in land improvement from 1880 to 1900. However,
it casts some doubt on whether previously estimated increases in farmland settlement should
indeed be attributed to these counties' lower levels of woodland.
(iv) Functional Form and the Exclusion of State-Decade F.E. The baseline estimates are not sensitive to using different polynomial functional forms to model the dependence of agricultural outcomes on woodland. As long as the functional form is sufficiently
flexible to capture the basic non-linearity appearing in figure 4, the evaluated effects at
benchmark woodland levels are similar.
Figure 6 displays the raw data underlying the main results, stripping away the regression
methodology. For each sample county, the change from 1880 to 1900 in the improvement
intensity of farmland is plotted against that county's 1880 woodland level. Counties with
the least woodland tended to experience large increases from 1880 to 1900, while counties
with more woodland were relatively unchanged.
The estimated changes from 1880 to 1900 decrease by roughly 20-40% when the stateby-decade fixed effects are replaced with decade fixed effects, which might be surprising if
one expected movement to low-woodland states in response to lower fencing costs. This is
entirely driven by Colorado, however, which is far from most previous settlement and has
low average woodland. When Colorado is dropped from the sample, the baseline results are
similar and the estimates increase 5-20% when the state-by-decade fixed effects are replaced
with decade fixed effects.
Summary of Robustness Tests
Throughout the different specifications discussed in these subsections, the baseline results are fairly robust for the estimated changes in the improvement intensity of farmland.
Counties with the least woodland generally experienced large relative increases in land improvement from 1880 to 1900. Similar increases did not occur before 1880, after 1900, and
did not occur at higher woodland levels from 1880 to 1900. Changes in total farmland are
less robust to alternative specifications, in particular, controlling for Westward settlement
and convergence in settlement.
V.B
Crop Productivity and Crop Choice
All cropland is "improved" by definition, so the previous estimates do not address whether
farmers adjusted crop production along the intensive margin. Given the secure protection of
property rights, the optimal allocation of land need not favor crops over livestock, as Coase
discusses. In the absence of secure property rights, however, the returns to certain crops
would likely be more sensitive to the threat of uncompensated damage by others' livestock.
Farmers might reduce investments in crop growing, harvest produce early, or otherwise adjust
production in ways that contribute to sub-optimal productivity. This section presents results
on changes in crop productivity, as a summary measure of all unobserved adjustments in
crop production.
Beginning in 1880, county-level data are available on the total production and acreage for
each of the six main crops on the Plains (corn, wheat, hay, oats, barley, rye). 40 Productivity
for each crop p in each county c is defined as its total production divided by the total acreage
devoted to that crop. For ease of interpretation, productivity in each decade is normalized
by its value in 1880.41 Measurement error creates some extreme outliers, so the upper and
lower centiles of the normalized productivity distribution are dropped.4 2
To implement the analysis, equation (8) is slightly modified: state-decade fixed effects
are replaced with crop-state-decade fixed effects, and the equation is first-differenced by
crop-county. For this baseline analysis, data from all crops are stacked so that the change
40
Cotton is excluded from the analysis, as data are only available for Texas and the boll weevil blight
severely impacted cotton productivity. Using the same technique as before, the data are adjusted to maintain
the 1880 geographical boundaries.
41
This gives similar results to analyzing changes in the log of productivity, but the estimates are always
relative to productivity in 1880.
42
This drops values less than 0.36 or greater than 6.4. The results are not sensitive to these cutoffs, as
long as the clearly extreme observations are dropped.
in productivity across woodland levels is constrained to be the same for all crops:
(9)
Yt
Ypc(t-1)
Ypc1880
=
pst
+
I
c +
2
W
3
+ 04tW+
ept
t=2
Columns 1 and 2 of Table 5 present baseline results, evaluated at the same representative
woodland levels as before. From 1880 to 1890, average productivity across all six crops
increased 23.4% more in a county with 0% woodland than in a county with 6% woodland.4 3
Over this same period, crop productivity increased by only 1.8% more in a county with
6% woodland than in a county with 12% woodland; a statistically smaller change with a tstatistic of 3.31. To give a sense of the large magnitude of this 23.4% increase, total US crop
production increased by roughly 1.7% annually from 1880 to 1920. US crop yields increased
by only 0.23% annually from 1880 to 1920, though this is likely to understate productivity
growth due to compositional changes in cropland. 44
In contrast to previous land-use results, these estimated increases did not continue
through the 1890s: crop productivity decreased by a statistically insignificant 4.6% from
1890 to 1900, leaving it 18.8% higher than in 1880. One interpretation is that the moreimmediate reaction to barbed wire's availability was to secure cropland. Even as barbed wire
became more widely used from 1890 to 1900, productivity would only increase if cropland
became more secure which, in turn, could only happen if farmers had continued to cultivate
substantial amounts of land without fences after 1890.
To explore the timing of these effects, equation (8) is estimated for the fraction of farmland
planted in crops.
Columns 3 and 4 of Table 5 report the results.
From 1880 to 1890,
the allocation of farmland to crops increased by 12 percentage points in a county with 0%
43
An alternative specification would weight observations by the number of county acres devoted to that
crop in 1880. This specification yields similar estimates, with the coefficient increasing from 23.4% to 26.9%.
The standard error changes only minimally, from 5.7% to 5.8%. Without differences in efficiency motivating
the weighting, it seems more transparent to focus simply on the unweighted results.
44
These numbers were estimated using indexes of total US crop production and yield per acre harvested
for twelve major crops (NBER Macrohistory Database, files a01005aa and a01297). The production index
was computed by weighting the production of each commodity by average farm prices from 1910-1914, and
both indexes are then defined relative to a base year. To obtain the average annual increase, the natural log
of each index is regressed on a time trend from 1880 to 1920.
woodland relative to a county with 6% woodland, with little change between higher woodland
levels. From 1890 to 1900, there was little change in the allocation of farmland to crops.
It appears that farmers' initial reaction (1880-1890) was mainly to expand crop production, both in terms of a greater intensity of cultivation and a greater fraction of farmland
allocated to crops. Then, from 1890 to 1900, increases in the improvement intensity of farmland were driven mainly by improvements to pasture land or land otherwise not currently
planted in crops.
Extensions and Robustness Tests
(i) At-Risk Crops vs. Hay. A natural extension of these results exploits crop-level
differences in vulnerability to livestock damage. While cattle eat hay (various grasses), fields
of hay are more resistant to livestock damage before being harvested. Fields of hay can even
be intended for grazing at certain times of the year. The other crops (corn, wheat, oats,
barley, rye) would yield substantially less grain if they were trampled, so it would be more
critical to protect them from others' livestock.45
Restricting the analysis to these five crops more at-risk of damage, columns 5-6 report
the results from estimating equation (9). From 1880 to 1890, productivity increased by 29%
in counties with the least woodland and the results are otherwise similar. By contrast, there
was no change in hay productivity from 1880 to 1890 (not shown). Columns 7-8 report
the results from estimating equation (8) for the fraction of cropland allocated to these more
at-risk crops. Indeed, from 1880 to 1890, more cropland became allocated to these five crops
instead of to hay.
(ii) Cropland Composition. Estimated changes in productivity could be confounded
by compositional changes in cropland, to the extent that farmland differs in productivity
within a county. This seems unlikely, however, to explain the observed large increases in
productivity from 1880 to 1890. The typical response would be to expand production into
45
Even controlled grazing would typically lower grain yields by 25-79% [Smith et al. 2004]. Wheat, oats,
barley, and rye could be grown for hay instead of for grain, but would then be managed differently. The
Census defines data for these crops as that which is grown for grain.
otherwise unprofitable and lower quality land. This would lead the empirical analysis to
understate the increase in productivity for a fixed plot of land. If production expanded into
less-wooded areas of counties that were inherently more productive, it could contribute to
the observed increases in productivity. This is likely of little substantive importance, though,
as very little of the land in farms was actually wooded: in 1880, less than 6% of the land
in farms was wooded in 76% of the counties with less than 6% woodland. The settlement
of new farmland is sometimes associated with short-term productivity benefits, but these
increases in productivity mostly persisted.
(iii) Regional Development.
In an attempt to account for potential region-specific
technological improvements, this subsection briefly reports on robustness checks using the
previously defined land resource regions, subregions, and soil groups. The estimated increases
in productivity from 1880 to 1890 are robust to the inclusion of crop-specific quadratic time
trends by region, subregion, or soil group.4 6 The estimated increase in cropland intensity
from 1880 to 1890 is robust to including quadratic time trends by region or soil group; by
subregion, the increase is reduced by one-third but is statistically significant.4
(iv) Westward Development.
7
Another potential concern is that there could be a
Westward trend associated with cropland management. The estimated increase in cropland
intensity from 1880 to 1890 is robust to controlling for distance West and distance from St.
Louis in each decade, with or without the inclusion of Colorado and Texas. Recall that
excluding Colorado and Texas forces the distance control variables to be estimated on the
same set of counties that mostly identify the changes associated with woodland levels.
When excluding Colorado and Texas, the crop productivity results are robust to including
the distance controls in each decade. The productivity results are sensitive, however, to these
control variables when including Colorado and Texas: there is no 1880 to 1890 increase and
46
When allowing for crop-decade fixed effects by region or soil group, the 1880 to 1890 increase is roughly
half its original size and marginally statistically significant. Productivity across woodland levels is also less
stable in subsequent periods. Notably, the results are not robust to controlling for crop-decade-subregion
fixed effects, though that is quite demanding of the data.
47
These results are also similar when replacing the quadratic trends with decade fixed effects by region,
soil group, or subregion.
productivity is quite variable in later periods. Given this difference in results, it seems
appropriate to place more emphasis on the results when the woodland variables of interest
and the distance control variables are primarily estimated on the same sample of observations.
Summary of Extensions and Robustness Tests
Counties with the least woodland appear to experience a substantial increase in crop
productivity following the introduction of barbed wire. These productivity increases are
concentrated among those crops most at-risk without fencing. Farmland was increasingly
allocated to crops and, in particular, more at-risk crops. These results are mostly robust
to the inclusion of regional control variables, but there are some notable exceptions: the
inclusion of crop-subregion-decade fixed effects, and controlling for distance West and from
St. Louis when the sample includes Colorado and Texas.
V.C
Land Value
One can gauge the monetary value of barbed wire to farmers, by estimating the resulting
change in land values. While land improvement and crop productivity may have increased
substantially, these changes would have come at considerable cost to farmers. Changes in
land values may capitalize the net benefit to agricultural production, if land markets were
well-functioning. In this era, there was a large amount of land speculation and taxes were
paid on the value of land [Gates 1973], so it appears plausible that farmers would have some
sense of their lands' value when asked by Census enumerators. 4 8
In each period, "land value" data are available for the combined value of farmland,
buildings, and fences. 49 However, land appears to have been the largest component of this
measure. In 1900 and 1910, the value of buildings averaged between 13% and 17% of the
total for low-, medium-, and high-woodland counties. Fence stock values are unknown, but
48
1f farmers were very quick to update reported valuations, they may have partly anticipated the arrival
of barbed wire by the 1880 Census. Unsettled land would still have been valued at zero, though, so much of
the response would still not occur until newly desirable lands were settled.
49
Additional data are obtained from Haines and ICPSR [2005].
the costs of building and repairing fences in 1879 averaged 1% of the total value of land,
buildings, and fences.
The natural log of land value is analyzed because technology, prices, and property rights
are typically modeled to have multiplicative effects on the total value of production. Additive
shocks would have a larger percentage effect in areas with low initial levels, so equation (8) is
estimated controlling for differences in initial land values. 50 For example, given similar large
increases in settlement across all counties from 1870 to 1880, land values would increase by
a larger percentage in low-woodland counties that averaged lower initial land values.
Table 6 presents the results. Land values increased substantially from 1880 to 1900 in
counties with 0% woodland relative 6% woodland, though only the change from 1880 to 1890
was statistically greater than the relative increases at higher woodland levels. By contrast,
before 1880 and after 1900, there were either relative declines or small changes.
Focusing on the more robust increase from 1880 to 1890, this represents an economically
substantial change. The 1880 to 1890 change of 0.406 log points is an increase of 50% above
1880 levels. This is approximately 1.7 times the 1880 value of all agricultural products in
low-woodland counties. If we assume that barbed wire had no effect on counties with more
than 6%woodland, these estimates imply that the overall monetary benefit of barbed wire to
farmers in sample counties was $103 million with a standard error of $32 million, in 1880 US
dollars.51 This is approximately 0.9% of total US GDP in 1880 [Historical Statistics 2006].
Throughout various robustness checks, this total estimated value ranges from $67 million to
$139 million and is always statistically significant. 52
50
Included is an additional fourth-degree polynomial function of log land value in 1870, interacted with
each decade. Without these additional controls, the estimates fit a clear pattern of economic convergence
for both low- and medium-woodland counties: there are large relative increases from 1870 to 1880 that
then decline over time. This specification is subject to the same concerns as before, but it adjusts for this
otherwise dominant pattern of convergence.
51
Sixty-four counties had between 0% and 1% woodland, with an average of 0.42% woodland and $1.7
million of land value in 1880. For an average county with 0.42% woodland, land values increased by an
estimated 43% relative to a county with 6% woodland. This gives an overall effect of approximately $47
million (64 x $1.7m x 43%). Summing across the woodland bins (1-2%, 2-3%, 3-4%, 4-5%, 5-6%) yields an
estimate of $103 million with a standard error of $32 million.
52
These include controlling for quadratic time trends or decade fixed effects by region, subregion, or soil
group; distance West by decade; distance West and distance from St. Louis by decade.
VI
Analysis of Potential Confounding Factors
VI.A
Other Shocks Correlated with Woodland Levels
The empirical results would not identify the effects of barbed wire to the extent that other
factors caused counties with less woodland to change differently from 1880 to 1900. As a
complement to the previous robustness checks, this section provides qualitative evidence on
other potential sources of differential changes.
One concern is that a shock to the timber market might have differentially affected
farming in counties with different amounts of woodland.5 3
However, the timber market
in all sample counties was a very small sector of the economy: in 1870, forest products
averaged only 0.6% of the total value of all farm products. These counties had such little
woodland that any changes in its use would have had limited effects on overall land-use.
Land improvement in the Plains did not generally involve clearing woodland because there
was already so much open land.
Two other potential confounding factors were the various land acts and the railroads.
Due to the lack of fencing materials, however, the 1862 "homestead law was a snare and a
delusion" [Webb, p.286]. Subsequent revisionary land acts were generally ill-suited to the
region, poorly enforced, and are thought to have played little role in the successful settlement
of the Plains [Webb, pp.424-431]. 54 Furthermore, most railroad construction in the sample
region had been completed by 1873, when the railroad industry faced a panic due to excessive
construction and low usage on the Plains.55 Webb summarizes:
We have noted that the agricultural frontier came to a standstill about 1850, and
that for a generation it made but little advance into the sub-humid region of the
Great Plains. It was barbed wire and not the railroads or the homestead law
53
Indeed, barbed wire may have lowered demand for timber by replacing wooden fences. Barbed wire
could also have increased demand for timber because of increases in settlement. Farmers would still have
required some timber for buildings and fence posts.
54
See Libecap [2006] for an analysis of land allocation policies in these and more western regions.
55
The construction of spur lines is quite endogenous to settlement, so it would be difficult to interpret a
formal analysis controlling for spur construction even if that data were available.
100
that made it possible for farmers to resume, or at least accelerate, their march
across the prairies and onto the Plains. [pp.316-317]
Other potential confounding factors are changes in technology. The post-Civil War period
saw many technological improvements and mechanization began to increase in the 1900s,
but the rise of barbed wire from 1880 to 1900 appears to coincide with a notable lack
of other major advances [Rasmussen 1962, Primack 1977].
Olmstead and Rhode discuss
biological innovations in wheat, but the 1880 to 1890 period does not stand out. Windmills
were used increasingly during this period to obtain water in areas that were often also low
in woodland. Knowledge of windmills had existed for a long time, however, and modern
industrialized production was widespread by the 1860s. Webb concludes that the increased
use of windmills was caused by the introduction of barbed wire and increased cultivation of
water-scarce areas [Webb, p.341, p.348]. For the sample region, note also that most conflict
with Native Americans had ended by this time.
VI.B
Evidence on Potential Other Effects of Barbed Wire
The empirical results would not identify only the effects of property rights if barbed wire had
important effects through other channels. Even if land were perfectly secure and there was
no threat of uncompensated damage by others' livestock, cheaper fencing might have some
effect on agricultural production. If fencing allowed greater control over cattle feeding and
breeding, it might increase the productivity of land for raising livestock. By enabling the
56
joint production of cattle and crops, fencing might also improve production possibilities.
If these effects were substantively important, then barbed wire should lead to an increase
in cattle production and/or the joint production of cattle and crops. The security of property
rights might also affect these outcomes, but these tests are designed to address whether nonproperty right effects could be driving the earlier results."5
56
Empirical estimates suggest
For example, if nearby lands varied in their suitability for cattle and crops.
If so, it seems reasonable to expect that non-property right effects should also dominate along these
dimensions.
57
101
that barbed wire did not increase cattle production and decreased the joint production of
cattle and crops. This finding is consistent with the notion that barbed wire mainly affected
agricultural outcomes through greater enforcement of property rights.
To test for increases in cattle production, equation (8) is estimated for the number of
cattle per five county acres.5" The estimated magnitudes can be roughly compared to previously estimated changes in land-use, as it requires approximately five acres of land for a
cow to graze in this region.
Columns 1 and 2 of Table 7 present the results.
From 1880 to 1890, there was no
substantial increase in cattle production for a county with 0% woodland relative to a county
with 6% woodland. The estimated coefficient implies that approximately 0.39% less of the
county land became devoted to cattle production. By contrast, cattle production increased
moderately in all subsequent periods and at higher woodland levels from 1880 to 1890. In
all, there is little indication that barbed wire specifically led to increased cattle production,
especially of a magnitude that could account for the previously estimated effects.
59
It is less clear how to estimate changes in the joint production of cattle and crops,
because the exact amount of land in each county devoted to cattle is unobserved. 60 As
a possible solution, I assume that all farmland not used for crops is either used for cattle
or other purposes that are not systematically changing over time within states. Based on
this assumption, I define a specialization index equal to the squared difference between the
percentage of county farmland devoted to crops and the average over all counties in that
decade and state: I' = (M - Mt) 2 . This index increases when a county with an aboveaverage percentage of farmland devoted to crops increases that proportion, and vice versa.
58Data on cattle are first available in 1880, so county regions are held constant at their 1880 boundaries.
59
These results are similar when controlling for quadratic time trends by region, region-decade fixed effects,
distance West by decade, or distance West and distance from St. Louis by decade. When controlling for
quadratic time trends or decade fixed effects by soil group or subregion, there are some increases in cattle
production from 1880 to 1890. These increases are never substantively or statistically greater than at higher
woodland levels, and the magnitudes are not greater than in later periods.
60
The assumption of five acres per cow is too rough for these purposes, as acreage requirements vary with
the environment, production methods, desired cattle quality, and desired sustainability.
102
Changes in this index are estimated using equation (8), which controls for average county
deviations from the mean and state-by-decade shocks.
Columns 3 and 4 of Table 7 present the results.
From 1880 to 1890, counties with
0% woodland became increasingly specialized, relative to counties with 6% woodland. The
estimated magnitude is difficult to interpret, but it is large relative to the index mean and
standard devation.
Throughout various robustness checks, this increase is either similar
or larger and always statistically significant. 6 1 Thus, it does not appear that barbed wire
encouraged the joint production of cattle and crops.
Overall, the empirical estimates fail to contradict the hypothesis that barbed wire's effects
operated mainly through an increase in property rights enforcement.
VII
Evidence on States' Legal Reforms
Prior to barbed wire's introduction, there were contentious political debates over whether
settlers should be granted full land protection without having to build a fence. In contrast
to the example of a farmer and cattle-raiser considered by Coase, individual cattle-raisers
would mainly have been unable to provide effective guarantees to the farmer.6 2
Plains
state legislatures considered passing "herd laws" that would make livestock owners formally
responsible for damages to farmers' unfenced crops.
These legislative debates highlight the property rights problem, but contentious debates
do not imply that legislative responses could be effective. "Herd laws" would only be effective
if enforced sufficiently to provide the credible promise of compensation if farmers' crops were
damaged. Webb argues that formal laws often had little practical influence on the Plains.
Table 8 reports the results from separately estimating equation (8) in Kansas and Nebraska, two states to implement versions of these reforms described by Kawashima and Davis
61
These include controlling for quadratic time trends or decade fixed effects by region, subregion, or soil
group; distance West by decade; distance West and distance from St. Louis by decade.
62
Because of the large number of current and potential cattle-raisers, transaction costs were prohibitive.
103
[1973].63 The empirical estimates are less precise within individual states, but the bulk of
the evidence indicates that the legal reforms were not effective.
Nebraska adopted a state-wide herd law in 1871, which should have benefited farmers
more in counties with less woodland and higher fencing costs. From 1870 to 1880, however,
the improvement intensity of farmland declined substantially in a county with 0% woodland
relative to a county with 6% woodland. It was not until 1890, after the introduction of
barbed wire, that counties with the least woodland showed substantial increases in land
improvement. Note that the coefficients are imprecisely estimated at higher woodland levels
because 90% of the sample counties in Nebraska had less than 6% woodland.
Kansas gave counties the option of adopting a herd law in 1872. I do not know exactly
which counties adopted the herd law, but it was generally not adopted in counties with little
woodland where the small proportion of crop growers had weak political influence [Webb,
p.500; Sanchez and Nugent 2000; Kawashima]. Settlement increased more in lower-woodland
counties from 1870 to 1880, however, and there was no change in land improvement.
It
was not until 1880 to 1890, after the introduction of barbed wire, that land improvement
increased in counties with the least woodland. Kansas then adopted a state-wide herd law in
1889, which should have benefited farmers in lower-woodland counties, but land improvement
declined from 1890 to 1900.
Davis argues that herd laws provided many of the benefits that were subsequently reinforced by the introduction of barbed wire, but most evidence comes from newspaper advocacies for legal reform. There is little evidence that the laws were strictly enforced and,
as one would expect, cattle owners resisted making compensating payments for damages.
Once barbed wire was introduced, those same newspapers wrote that "every farm needs
some fencing" and as "soon as a farmer is able, he fences his farm. There must be an apparent benefit" [Davis, p.134]. Farmers made substantial investments in fencing before the
legal reforms, after the legal reforms, and after the introduction of barbed wire. These laws
63
Texas also adopted a county-option policy in 1870, but the results are mostly noise because the sample
has only a small number of counties in Texas with little woodland.
104
may have had some small influence or they might not have been so hotly debated, but their
passage appears to reflect changes in local political power rather than a major influence on
farmers' decisions. In the absence of physical barriers, formal laws appear to have provided
farmers little refuge from roaming livestock.
VIII
Conclusion
In the
century American Plains, fences were required to secure farmland. Initially, fences
1 9 th
were prohibitively expensive in areas with little local woodland. In this historical context,
the introduction of barbed wire fencing encouraged property rights protection most in areas
with the least woodland. This paper tests the hypothesis that this increase in property rights
protection played an important role in the agricultural development of the American Plains.
Between barbed wire's introduction and its universal adoption, counties with the least
woodland are estimated to have experienced substantial relative increases in agricultural
development. Increases along intensive margins appear to have been dominant: farmland
became more intensely improved, a greater percentage of farmland was allocated to crops,
and crop productivity increased. Estimated increases in land values suggest that these effects
were associated with substantial economic gains. On balance, these findings appear to be
robust to a range of specification tests.
Overall, the results here suggest that secure property rights over land may have an
important role in facilitating agricultural development.
In contrast to the typical policy
reform, however, these results reflect changes in property rights that largely fall outside
formal institutional channels. Indeed, the apparent ineffectiveness of contemporaneous legal
reforms may indicate a difficulty protecting property rights in sparsely settled areas through
legislation and the court system.
105
Data Appendix
The data was converted into time-invariant geographical units using ArcView (GIS) software
and the Historical United States Boundary Files [Carville et al. 1999].64 The conversion
process is designed to obtain data for each 1870 county when the data are subsequently
reported for different county definitions. This turns out not to be particularly important
for the empirical results, because most counties that change borders are excluded from the
sample for other reasons (Figure 2). Still, this process is qualitatively important for the data
to make sense and to give each original 1870 county equal weight.
This process is now described for converting 1920 data into the 1870 county definitions.
The 1870 map is intersected with the 1920 map and each 1920 county is assigned the identification number of its 1870 county origin. When the 1920 county falls within more than
one 1870 county, each piece of the 1920 county is assigned its unique 1870 origin along with
the area of that piece. This information is then matched to the Census data, where the
adjusted data for each 1920 piece are found by multiplying the original 1920 Census value
by that piece's relative size. Finally, the 1920 data for each 1870 county region are found by
summing across all 1920 pieces that make up the region. 65
When counties in later periods fall within more than one original 1870 region, it is
impossible to assign the exact measure of the variable in that period to each original region.
This procedure assumes that each variable is evenly distributed across the county area.
This introduces no additional measurement error for the typical changes in county borders
over this time, whereby a single county split into subsections. In practice, the digital map
projections are imperfect, resulting in small sections of counties being assigned to different
original regions, but the effect on the total variable measures is very small. For the six states
of interest, roughly 85% of counties in every period have less than 1% of their area in a
second original region.
64
The map files are available at the beginning of each decade. In each period, 3-9 census counties are
dropped because they do not match to the map files.
65
The summed measure is considered to be missing when data for any piece is missing.
106
In the final data, counties will not have the exact same number of acres in different
periods. Slight changes in the map projections or data collection procedures would create
small differences. There are some very large differences, by hundreds of thousands or millions
of acres, that must be due to data coding errors or mismatches regarding when large border
changes took place. Counties for which the final standard deviation of the number of acres
is greater than 50,000 are dropped (3% of the sample).
107
References
Acemoglu, Daron and Johnson, Simon. 2005. "Unbundling Institutions." Journalof Political
Economy. Vol. 113, No. 5, pp. 949-995.
Alston, Lee J., Gary D. Libecap, and Bernardo Mueller. 1998. "Property Rights and Land Conflict:
A Comparison of Settlement of the U.S. Western and Brazilian Amazon Frontiers," in Latin
America and the World Economy Since 1800, ed. John H. Coatsworth and Alan M. Taylor.
Cambridge: Harvard University Press: pp. 55-84.
Anderson, Terry L. and Hill, Peter J. 1975. "The Evolution of Property Rights: A Study of the
American West." Journalof Law and Economics. Vol. 18, No. 1, pp. 163-179.
Banerjee, Abhijit V.; Gertler, Paul J.; Ghatak, Maitreesh. 2002. "Empowerment and Efficiency:
Tenancy Reform in West Bengal." Journal of PoliticalEconomy. Vol. 110, No. 2, pp. 239-280.
Besley, Timothy. 1995. "Property Rights and Investment Incentives: Theory and Evidence from
Ghana." JournalofPoliticalEconomy. Vol. 103, No. 5, pp. 903-937.
Bogue, Allan G. 1963a. From Prairie to Corn Belt: Farming on the Illinois and Iowa Prairies in
the Nineteenth Century. University of Chicago Press.
Bogue, Allan G. 1963b. "Farming in the Prairie Peninsula, 1830-1890." JournalofEconomic
History. Vol. 23, No. 1, pp. 3-29.
Brasselle, Anne-Sophie; Gaspart, Frederic; Platteau, Jean-Philippe. 2002. "Land Tenure Security
and Investment Incentives: Puzzling Evidence from Burkina Faso." Journalof Development
Economics. Vol. 67, No. 2, pp. 373-418.
Carville, Earle; Heppen, John; Otterstrom, Samuel. 1999. HUSCO 1790-1999: historical United
States county boundary files [electronic resource]. Louisiana State University (Baton Rouge,
LA): Geoscience. http://www.ga.lsu.edu/husco.html
Cheung, Steven N.S. 1970. "The Structure of a Contract and the Theory of a Non-Exclusive
Resource." Journalof Law andEconomics. Vol. 13, No. 1, pp. 49-70.
Coase, Ronald H. 1960. "The Problem of Social Cost." Journalof Law and Economics. Vol. 3,
No. 1,pp. 1-44.
Davis, Rodney 0. 1973. "Before Barbed Wire: Herd Law Agitations in Early Kansas and
Nebraska," in Essays in American History in Honor of James C. Malin, ed. Burton J. Williams.
Coronado Press.
Demsetz, Harold. 1967. "Toward a Theory of Property Rights." American Economic Review.
Papersand Proceedings. Vol. 57, No. 2, pp. 347-359.
Encyclopedia Britannica, eleventh edition. 1911. "Barbed Wire."
http://1911 encyclopedia.org/B/BA/BARBED_WIRE.htm
Gates, Paul W. 1973. "The Role of the Land Speculator in Western Development," in Landlords
and Tenants on the Prairie Frontier, pp. 48-71. Cornell University Press.
Goldstein, Markus and Udry, Christopher. 2005. "The Profits of Power: Land Rights and
Agricultural Investment in Ghana." Mimeo, Yale University, November.
Gutmann, Myron P. 2005. "Great Plains Population and Environment Data: Agricultural Data
[Computer file]." Ann Arbor, MI: University of Michigan / Ann Arbor, MI: ICPSR.
Hayter, Earl W. 1939. "Barbed Wire Fencing - A Prairie Invention." AgriculturalHistory. Vol.
13, pp. 189-207.
Haines, Michael R., and the Inter-university Consortium for Political and Social Research. 2005.
Historical, Demographic, Economic, and Social Data: The United States, 1790-2000 [Computer
file]. ICPSR02896-v2. Hamilton, NY: Colgate University / Ann Arbor, MI: ICPSR.
Jacoby, Hanan G.; Li, Guo; Rozelle, Scott. 2002. "Hazards of Expropriation: Tenure Insecurity
and Investment in Rural China." American Economic Review. Vo. 92, No. 5, pp. 1420-1447.
108
Kawashima, Yasuhide. 1994. "Fence Laws on the Great Plains, 1865-1900," in Essays on English
Law and the American Experience, eds. Elizabeth A. Cawthorn and David E. Narrett. Walter
Prescott Webb Memorial Lectures, No. 27, pp. 100-119. Texas A&M University Press.
Kraenzel, Carl Frederick. 1955. The Great Plains in Transition. University of Oklahoma Press.
Libecap, Gary D. 2006. "The Assignment of Property Rights on the Western Frontier: Lessons for
Contemporary Environmental and Resource Policy." NBER Working Paper 12598.
McFadden, Joseph M. 1978. "Monopoly in Barbed Wire: The Formation of the American Steel
and Wire Company." Business History Review, Vol. 52, No. 4, pp. 465-489.
McCallum, Henry D. and McCallum, Frances T. 1965. The Wire that Fenced the West. University
of Oklahoma Press: Norman. 1972 printing.
Miceli, Thomas J.; Sirmans, C.F.; Kieyah, Joseph. 2001. "The Demand for Land Title
Registration: Theory with Evidence from Kenya." American Law and Economics Review. Vol.
3, No. 2, pp. 275-287.
North, Douglass C. 1981. Structure and Change in Economic History. W.W. Norton & Company.
Olmstead, Alan L. and Paul W. Rhode. 2002. "The Red Queen and the Hard Reds: Productivity
Growth in American Wheat, 1880-1940." Journal of Economic History. Vol. 62, No. 4, pp.
929-966.
Primack, Martin L. 1966. "Farm Capital Formation as a Use of Farm Labor in the United States,
1850-1910." Journalof Economic History. Vol. 26, No. 3, pp. 348-362.
Primack, Martin L. 1969. "Farm Fencing in the Nineteenth Century." Journalof Economic
History. Vol. 29, No. 2, pp. 287-291.
Primack, Martin L. 1977. Farm Formed Capital in American Agriculture 1850-1910. 1962
Dissertation. Arno Press: New York.
Rasmussen, Wayne D. 1962. "The Impact of Technological Change on American Agriculture,
1862-1962." Journalof Economic History. Vol. 22, No. 4, pp. 578-591.
Sanchez Nicolas and Nugent, Jeffrey B. 2000. "Fence Laws vs. Herd Laws: A NineteenthCentury Kansas Paradox." Land Economics. Vol. 76, No. 4, pp. 518-533.
Smith, S. Ray; Benson, Brinkley; Thomason, Wade. 2004. "Growing Small Grains for Forage in
Virginia." Virginia Cooperative Extension, Publication No. 424-006.
U.S. House of Representatives, 42 nd Congress, 1 st session. 1871. "Special Report on Immigration;
accompanying Information for Immigrants..." by Edward Young, Chief Bureau of Statistics.
Washington: Government Printing Office.
U.S. House of Representatives, 4 2 nd Congress, 2 nd session. 1872. "Statistics of fences in the United
States." House Executive Documents. Vol. 17, No. 327 (Report of the Commissioner of
Agriculture), pp. 497-512. Washington: Government Printing Office. (US Congressional
Serial Set Vol. 1522).
Webb, Walter Prescott. 1931. The Great Plains. Grosset & Dunlap: New York. 1978 printing.
109
Table 1. Statistics on Barbed Wire Production, Fence Stocks, and New Fence Construction
A. Annual Production. thousands of tons
1874
1875
1911 Encyclopedia
Britannica
0.005
0.3
1874
1875
Webb 1931, p.309
0.005
0.3
Hayter 1939
1876
1.5
1876
1.3
1877
7
1877
6
1878
13
1878
12
1880
1879
25
40
1880
1879
23
37
1880-1884
80-100
1890
125
1888
150
1900
200
1901
- 250
1895
157
1907
250
B. Fence Stocks, millions of rods (1 rod = 5 meters)
North Central
1850
1860
1870
1880
1890
1900
1910
Total
228
303
359
427
443
493
483
Wood
226
285
320
369
279
192
75
Stone
2
3
3
3
0
0
0
27
27
30
22
26
9
0
Hedge
Wire
0
6
14
30
137
271
382
South Central
Total
175
230
245
344
531
685
701
Wood
171
219
235
330
425
411
280
Stone
3
5
4
7
0
0
0
Hedge
0
5
2
3
0
0
0
0
2
3
3
106
274
420
Wire
Prairie
Total
5
22
41
80
255
607
718
Wood
4
17
23
40
130
176
7
Stone
0
1
2
2
0
0
0
26
3
18
22
0
3
13
Hedge
Wire
0
1
4
12
122
413
689
Southwest
Total
39
78
94
162
280
710
749
Wood
38
71
80
123
174
312
187
Stone
1
2
2
2
0
0
0
Hedge
0
2
4
5
0
0
0
Wire
0
4
9
32
106
398
562
C. New Fence Construction, percentage
North Central
1850-59
1860-69
1870-79
1880-89
1890-99
1900-09
Wood
79
66
73
3
0
0
Stone
1
0
0
0
0
0
Hedge
12
21
6
1
0
0
Wire
8
13
22
96
100
100
South Central
Wood
90
94
100
50
0
0
Stone
2
0
0
0
0
0
Hedge
4
1
1
0
0
0
3
5
0
50
100
100
Wire
Prairie
Wood
71
39
38
45
18
0
Stone
5
4
0
0
0
0
Hedge
18
45
38
0
0
0
Wire
6
12
24
55
82
100
Southwest
Wood
84
56
63
42
32
0
Stone
2
3
0
0
0
0
Hedge
5
11
2
0
0
0
Wire
9
29
35
58
68
100
Notes: "Wood" fences include three types: Virginia worm, post and rail, and board. "Wire" fences are the smooth iron
variety from 1850-1870 and include barbed wire from 1880 on. Panel B is excerpted from Primack 1977, table 23, pp.
206-208. Panel C is excerpted from Primack 1977, table 26, pp. 83-84.
110
Table 2. Mean County Characteristics in 1880, by County Woodland Group
1870 county boundaries
# counties
Land-use outcomes:
Acres of land in farms,
per county acre
Acres of improved land,
per acre in farms
Other characteristics:
Acres in county
Acres of land in farms
Acres of improved land
Value of land, buildings,
and fences
Value of all products
Cost of building and
repairing fences
1880 county boundaries
# counties
Acres of cropland,
per acre in farms
Acres of cropland
All
Counties
(1)
Low
Woodland,
0% - 4%
(2)
Medium
Woodland,
4% - 8%
(3)
High
Woodland,
8% - 12%
(4)
P-value
(2) vs. (3)
(5)
P-value
(3) vs. (4)
(6)
377
147
57
43
--
--
0.53
(0.26)
0.54
(0.23)
0.42
(0.26)
0.55
(0.20)
0.59
(0.28)
0.64
(0.28)
0.65
(0.26)
0.65
(0.24)
0.000
0.257
0.032
0.835
550,718
(526,638)
237,407
(113,987)
133,995
(91,561)
3,192,401
(2,851,257)
838,986
(659,730)
33,514
(24,120)
645,898
(801,941)
188,967
(125,574)
108,404
(82,723)
2,294,290
(1,750,834)
593,556
(452,311)
23,267
(18,173)
470,219
(239,127)
242,318
(104,289)
170,818
(110,841)
4,271,744
(3,635,498)
1,060,637
(867,683)
41,589
(31,547)
430,631
(180,333)
254,210
(92,210)
177,500
(98,161)
4,776,354
(3,245,953)
1,158,761
(756,448)
40,162
(19,616)
0.018
0.348
0.002
0.548
0.000
0.751
0.000
0.467
0.000
0.548
0.000
0.782
490
246
61
44
--
--
0.31
(0.20)
71,436
(65,041)
0.29
(0.19)
54,350
(54,606)
0.38
(0.25)
102,594
(85,283)
0.42
(0.19)
114,472
(73,867)
0.009
0.415
0.000
0.448
% acreage for each crop:
Corn
40.2
34.3
51.1
42.6
0.000
0.060
(22.4)
(23.6)
(24.0)
(21.3)
Wheat
23.2
28.5
22.6
24.1
0.033
0.711
(19.6)
(19.0)
(19.5)
(19.7)
Hay
18.3
26.2
13.7
14.6
0.000
0.662
(20.5)
(24.5)
(9.2)
(10.8)
Oats
7.6
8.2
6.7
8.6
0.050
0.037
(6.5)
(7.9)
(4.3)
(4.6)
0.589
0.664
1.0
1.4
1.2
1.0
Barley
(2.6)
(1.6)
(1.7)
(1.8)
0.6
0.5
0.773
0.499
Rye
0.5
0.6
(0.9)
(1.0)
(0.9)
(0.7)
Notes: In the top panel, the sample is the same as in Figure 2 and Table 3 (377 counties in a balanced panel). In the
bottom panel, the sample is the same as in Tables 5 and 7 (490 counties in a balanced panel). Missing data for crop
acreage is treated as a zero. P-values are calculated based on standard errors that are adjusted for heteroskedasticity.
111
Table 3. Estimated Changes in the Improvement Intensity of Farmland and Settlement,
Evaluated at Representative Woodland Levels
Woodland:
Before
Decade:
1870- 1880
Acres of Improved Land,
per acre in farms
0% vs. 6%
6% vs. 12%
(2)
(1)
0.015
(0.040)
0.023
(0.013)
[0.22]
After
Barbed Wire's
Introduction
Barbed
Wire's
Barbed
Wire's
Universal
Adoption
Acres of Land in Farms,
per county acre
0% vs. 6%
6% vs. 12%
(4)
(3)
0.039
(0.026)
[0.41]
0.028**
(0.010)
1880- 1890
0.100**
(0.030)
[4.12]
-0.004
(0.012)
0.129**
(0.023)
[3.81]
0.048**
(0.009)
1890- 1900
0.086**
(0.020)
[3.40]
0.020*
(0.009)
0.128**
(0.026)
[3.13]
0.057**
(0.009)
1900- 1910
- 0.019
(0.011)
[2.25]
0.004
(0.006)
- 0.022
(0.024)
[0.36]
-0.014
(0.009)
1910 - 1920
0.003
(0.010)
0.006
(0.004)
0.024
(0.016)
[1.59]
- 0.003
(0.007)
[0.31]
R2
0.4432
0.5012
Observations
1885
1885
Notes: For the indicated outcome variable, this table reports estimates from equation 8 in the text. The estimated
polynomial function of woodland is evaluated at three woodland levels (0%, 6%, 12%) and the reported estimates
represent the difference in predicted changes between the two indicated woodland levels. Each cell can be interpreted
as a difference-in-difference coefficient: the change over that period for an average county with 0% woodland relative
to the change for a county with 6% woodland (or a county with 6% woodland relative to a county with 12% woodland).
Robust standard errors clustered by county are reported in parentheses: ** denotes statistical significance at 1% and *
at 5%. In brackets is the t-statistic of the difference between the two contemporaneous estimates comparing changes at
different woodland levels.
112
Table 4. Estimated 1880-1890 and 1890-1900 Changes for 0% Woodland Counties vs. 6% Woodland Counties,
Robustness to Different Specifications
Baseline
Specification
Distance West
(2)
(1)
Panel A: Acres of Improved Land, per acre in farms
Distance West
and Distance
From St. Louis
(3)
Additional Controls for:
Quadratic Time
Quadratic Time
Trends by
Trends by
Region
Subregion
(5)
(4)
Quadratic Time
Trends by
Soil Group
(6)
1870 Outcome
Differences
(7)
1880-1890
0.100**
(0.030)
[4.12]
0.094**
(0.034)
[3.76]
0.119**
(0.031)
[4.54]
0.100**
(0.029)
[3.76]
0.087**
(0.031)
[3.29]
0.120**
(0.031)
[4.36]
0.065*
(0.033)
[2.90]
1890- 1900
0.086**
(0.020)
[3.40]
0.094**
(0.026)
[3.33]
0.106**
(0.026)
[3.64]
0.087**
(0.020)
[3.22]
0.070**
(0.021)
[2.61]
0.094**
(0.021)
[3.50]
0.076**
(0.021)
[2.95]
0.4589
1885
0.4773
1885
0.5095
1885
0.4864
1885
0.5735
1885
0.068*
(0.027)
[1.95]
0.083**
(0.026)
[2.33]
0.106**
(0.025)
[2.48]
0.078**
(0.024)
[1.87]
0.121**
(0.025)
[3.54]
0.031
(0.029)
[1.26]
0.079*
0.066*
(0.032)
[1.27]
0.116**
(0.026)
[2.48]
0.097**
(0.027)
[1.92]
0.122**
(0.026)
[2.91]
0.062
(0.039)
[0.99]
R2
0.4432
0.4458
Observations
1885
1885
Panel B: Acres of Land in Farms, per county acre
1880- 1890
1890-1900
0.129**
(0.023)
[3.81]
0.128**
(0.026)
[3.13]
(0.031)
[1.56]
R2
0.5012
0.5583
0.5716
0.5418
0.5737
0.5352
0.5606
1885
Observations
1885
1885
1885
1885
1885
1885
Notes: This table reports on a series of specification tests for the results in Table 3. Only the primary results are shown here: the estimated changes from 1880-1890
and 1890-1900 for counties with 0% woodland relative to counties with 6% woodland. The full results are shown in Appendix Tables 1-3. For comparison, the results
from Table 3 are reported in column 1. Column 2 controls for a county's distance West, interacted with each decade. Column 3 controls for a county's distance West
and distance from St. Louis, interacted with each decade. Column 4 controls for a quadratic time trend for each of 11 Regions. Column 5 controls for a quadratic time
trend for each of 43 Subregions. Column 6 controls for a quadratic time trend for each of 19 Soil Groups. Column 7 controls for a fourth-degree polynomial of a
county's 1870 outcome level, interacted with each decade.
113
Table 5. Estimated Changes in Crop Productivity and Crop Intensity
All Crops
Productivity
Woodland:
Decade:
1880-1890
After
Barbed Wire's
Introduction
After
BarbedWire's
Universal
Universal
0% vs. 6%
(1)
0.234**
(0.057)
[3.31]
6% vs. 12%
(2)
0.018
(0.025)
1890 -1900
-0.046
(0.039)
[0.97]
-0.003
(0.018)
1900 -1910
0.054
- 0.019
1910 -1920
Acres of Cropland,
per acre in farms
0% vs. 6%
6% vs. 12%
(3)
(4)
0.121**
(0.015)
[8.01]
-0.013
(0.009)
[0.24]
0.015**
(0.005)
At-Risk Crops
Acres of At-Risk Crops,
per acre of cropland
6% vs. 12%
0% vs. 6%
6% vs. 12%
(6)
(7)
(8)
Productivity
0% vs. 6%
(5)
0.292**
(0.067)
[3.47]
0.036
(0.027)
0.058*
(0.023)
[2.84]
0.000
(0.007)
- 0.011"*
(0.004)
-0.057
(0.046)
[0.86]
-0.012
(0.021)
0.007
(0.015)
[0.91]
-0.005
(0.006)
0.039**
0.004
0.058
- 0.027
0.020
(0.019)
[1.75]
- 0.015
(0.009)
0.016
(0.015)
[0.63]
0.005
(0.009)
(0.035)
[1.70]
(0.018)
(0.010)
[3.44]
(0.004)
(0.039)
[1.81]
(0.019)
-0.036
(0.042)
0.014
(0.025)
0.010
(0.007)
0.001
(0.004)
0.011
(0.049)
0.038
(0.029)
[0.95]
[1.06]
[0.44]
R2
0.3949
0.3922
0.4000
0.2777
Observations
9104
1960
7320
1960
Notes: For changes in productivity, estimates are from equation 9 in the text. For changes in cropland, estimates are from equation 8 in the text. Estimated
coefficients are presented in the same form as in Table 3.
114
Table 6. Estimated Changes in Land Value
Woodland:
Value of Land in Farms,
per county acre
0% vs. 6%
6% vs. 12%
(2)
(1)
Decade:
Before
Barbed Wire
After
Barbed Wire's
Introduction
After
Barbed Wire's
Universal
Adoption
1870 - 1880
S(0.151)
- 0.224**
- 0.364*
(0.053)
[1.11
[1.11]
1880 - 1890
0.406**
(0.105)
[3.99]
0.074*
(0.037)
1890 - 1900
0.213**
(0.072)
[1.45]
0.126**
(0.027)
1900- 1910
- 0.101
(0.073)
[2.07]
0.013
(0.027)
1910-1920
0.044
(0.063)
[0.50]
0.018
(0.024)
R2
0.7569
Observations
1880
Notes: Estimates are from equation 8 in the text and are reported in the same form as in Table 3.
115
Table 7. Estimated Changes in Cattle Production and County Specialization
In 1880:
Mean
Std. deviation
Woodland levels:
Decade:
1880 - 1890
After
Barbed Wire's
Introduction
Barbed Wire's
Universal
Adoption
Number of Cattle,
per five county acres
Degree of Specialization in
Crops or Cattle
0.2034
0.1570
0% vs. 6%
6% vs. 12%
(2)
(1)
0.0175
0.0248
0% vs. 6%
6% vs. 12%
(4)
(3)
- 0.0039
(0.0137)
[2.01]
0.0247**
(0.0066)
0.0117**
(0.0039)
[3.50]
- 0.0007
(0.0012)
0.0355**
(0.0071)
0.0007
(0.0032)
[2.05]
- 0.0070**
(0.0020)
1890- 1900
0.0567**
(0.0150)
[1.29]
1900 - 1910
0.0437**
(0.0140)
- 0.0122
(0.0065)
- 0.0023
(0.0025)
0.0021
(0.0014)
1910 - 1920
0.0348**
(0.0102)
- 0.0064
(0.0052)
- 0.0015
(0.0020)
[0.28]
- 0.0023
(0.0014)
[3.32]
R2
0.5823
0.1681
Observations
1960
1960
Notes: For the indicated outcome variable, estimates are from equation 8 in the text and are reported in the same form as in
Table 3.
116
Table 8. Estimated Changes in Improvement Intensity and Settlement, Nebraska and Kansas
Woodland:
Panel A: Nebraska
1870 - 1880
1880 - 1890
Acres of Improved Land,
per acre in farms
0% vs. 6%
6% vs. 12%
(2)
(1)
- 0.216*
(0.104)
0.294**
(0.062)
Acres of Land in Farms,
per county acre
0% vs. 6%
6% vs. 12%
(4)
(3)
0.301
(0.367)
0.034
(0.174)
0.640
(0.742)
- 0.350
(0.299)
0.060
(0.063)
0.206
(0.189)
- 0.120
(0.239)
1890 - 1900
- 0.069*
(0.033)
0.335*
(0.161)
1900 -1910
0.063
(0.035)
- 0.339
(0.181)
- 0.008
(0.043)
0.220
(0.148)
1910 - 1920
- 0.001
(0.026)
0.014
(0.075)
0.033
(0.019)
- 0.011
(0.107)
R2
Observations
Panel B: Kansas
1870 -1880
1880 -1890
0.175**
(0.057)
0.8129
155
- 0.027
(0.163)
0.307**
(0.060)
0.7534
155
0.011
(0.045)
0.479**
(0.102)
0.053*
(0.026)
- 0.135*
(0.066)
0.045
(0.035)
0.116**
(0.030)
1890 - 1900
- 0.091
(0.051)
0.008
(0.028)
0.017
(0.031)
0.027
(0.024)
1900 -1910
0.017
(0.041)
- 0.018
(0.026)
0.080*
(0.037)
- 0.023
(0.030)
1910 -1920
0.028
(0.030)
- 0.018
(0.015)
0.037
(0.019)
- 0.029*
(0.013)
R2
0.5283
0.7959
Observations
255
255
Notes: Panels A and B report the estimated effects from Table 3 for Nebraska only and Kansas only, respectively.
117
Figure 1. Declining Steel Prices and the Introduction of Barbed Wire
o
L-
0
LO
Cd-
o-
t-
0
-o.
. ....
..
Iron
-n
I
I
---------
Steel
S
x
l
Barbed
Wire
x Barbed Wire
approximately $450 per ton. Before 1890, "Steel" is the price of Bessemer steel rails in PA. After 1890, "Steel" is the
price of Bessemer steel billets in PA. "Iron" is the price of pig iron in PA.
118
Figure 2. Sample Counties by Defined Woodland Levels
Percent Woodland
o0%-1%
O
H
1%-2%
IO 2%-4%
4%-6%
6%- 8%
.
8%- 10%
10%-12%
12%- 14%
S14%+
Eli
O
SNot in Sample
Notes: This figure displays the 377 sample counties based on 1870 geographical boundaries, shaded to represent their
level of woodland.
119
Figure 3. Acres of Improved Land (per farm acre), by County Woodland Group and Decade
cl
1870
1880
0% to4%
1890
---
1900
4% to 8%
1910
1920
--------- 8% to 12%
Notes: For counties in these three woodland groups, this figure displays the average number of improved acres of land per
acre of farmland in each time period. The two vertical dotted lines represent the approximate date of barbed wire's
introduction to farmers (1880) and barbed wire's universal adoption by farmers (1900).
120
Figure 4. Estimated Changes in the Improvement Intensity of Farmland, +/- 2 Standard
Errors, Relative to a County with 0% Woodland
From 1880 To 1890
From 1870 To 1880
10
0
0
Co U.
(N
/
\
;
7
i
/
+o
ca
CU
\
01'
2!
CU.
(U
E
wE
U
(n
O
LO
LU
LU
(n
.N
I
CI
0
.02
.04
.06
.08
1
0
.12
.02
Woodland Level
.1
.1'2
From 1900 To 1910
Lo
From 1890 To 1900
.04
.06
.08
Woodland Level
+o
(D
From 1900 To 1910
W
LU
(I)
L+o
C
Ui
C4
I
-L
0
.02
.04
.06
.08
Woodland Level
.1
.12
0
.02
.08
.04
.06
Woodland Level
.1
.12
(0
+-
a5
C
r
(U
C
(U
01
>1
w
E
E
0
.02
.08
.04
.06
Woodland Level
.1
.12
Notes: From estimating equation 8 for the acres of improved land per acre of farmland, these figures report the change in a
county with each woodland level relative to the change for a county with 0% woodland, plus or minus 2 standard errors.
121
Figure 5. Estimated Cumulative Change in the Improvement Intensity of Farmland
C()
N*
.-
1870
0
1880
1890
1900
0 0% Woodland vs. 6% Woodland
1910
1920
0 6% Woodland vs 12% Woodland
Notes: This figure displays the cumulative estimated changes in acres of improved land per acre of farmland, as reported
in Table 3, columns 1 and 2. The two vertical dotted lines represent the approximate date of barbed wire's introduction to
farmers (1880) and barbed wire's universal adoption by farmers (1900).
122
Figure 6. County Changes in the Improvement Intensity of Farmland, 1880 to 1900
i
Q,*
0
sOI
* *o*
0D
0-
0
*00*
.*
0
*
0
*
*
*o*
0&
*
0
L
_
0
008
0
s0
I
1
.2
.3
Woodland
I
I
.4
.5
Notes: This figure shows the change from 1880 to 1900 in acres of improved land per acre of farmland, where each point
represents a single county. This change is plotted against that county's 1880 woodland level.
123
Appendix Table 1. Estimated Changes in the Improvement Intensity of Farmland and Settlement,
Controlling for Westward Development
Woodland:
Decade:
1870- 1880
Distance West
Improved Land,
Land in Farms,
per acre in farms
per county acre
0%-6%
6%- 12%
0%-6%
6%- 12%
(2a)
(2b)
(2c)
(2d)
-0.006
(0.044)
[0.55]
0.016
(0.013)
- 0.050
(0.024)
[2.02]
- 0.003
(0.009)
Distance West and from St. Louis
Improved Land,
Land in Farms,
per acre in farms
per county acre
0%-6%
6%- 12%
0%-6%
6%- 12%
(3a)
(3b)
(3c)
(3d)
- 0.005
(0.046)
[0.52]
0.016
(0.013)
- 0.042
(0.025)
[1.82]
0.001
(0.009)
1880 -1890
0.094**
(0.034)
[3.76]
- 0.007
(0.014)
0.068*
(0.027)
[1.95]
0.026**
(0.010)
0.119**
(0.031)
[4.54]
0.006
(0.013)
0.083**
(0.026)
[2.33]
0.033**
(0.009)
1890- 1900
0.094**
(0.026)
[3.33]
0.023
(0.012)
0.079*
(0.031)
[1.56]
0.040**
(0.011)
0.106**
(0.026)
[3.64]
0.029*
(0.011)
0.066*
(0.032)
[1.27]
0.034**
(0.011)
1900 -1910
- 0.009
(0.016)
[1.42]
0.008
(0.008)
- 0.009
(0.023)
[0.01]
- 0.009
(0.010)
- 0.015
(0.015)
[1.75]
0.005
(0.008)
0.007
(0.024)
[0.34]
- 0.001
(0.009)
1910 - 1920
0.006
(0.013)
[0.08]
0.007
(0.005)
0.033
(0.017)
[1.95]
0.000
(0.008)
0.003
(0.014)
[0.22]
0.006
(0.004)
0.023
(0.019)
[1.53]
- 0.005
(0.007)
R2
0.4458
0.5583
Obs.
1885
1885
Notes: This Table presents the full results from Table 4, columns 2 and 3.
0.4589
1885
124
0.5716
1885
Appendix Table 2. Estimated Changes in the Improvement Intensity of Farmland and Settlement,
Controlling for Regional Changes
Quadratic Time Trends by Region
Improved Land,
Land in Farms,
per county acre
per acre in farms
6% - 12%
0% - 6% 6% - 12%
0% - 6%
(4b)
(4c)
(4d)
(4a)
Quadratic Time Trends by Subregion
Improved Land,
Land in Farms,
per county acre
per acre in farms
0% - 6%
6% - 12%
0% - 6% 6% - 12%
(5b)
(5c)
(5d)
(5a)
0.013
(0.040)
[0.73]
0.039**
(0.013)
0.006
(0.024)
[1.22]
0.034**
(0.010)
0.004
(0.035)
[0.82]
0.031**
(0.012)
1880- 1890
0.100**
(0.029)
[3.76]
0.006
(0.011)
0.106**
(0.025)
[2.48]
0.052**
(0.009)
0.087**
(0.031)
[3.29]
0.000
(0.012)
1890- 1900
0.087**
(0.020)
[3.22]
0.025**
(0.009)
0.116**
(0.026)
[2.48]
0.060**
(0.010)
0.070**
(0.021)
[2.61]
0.019*
(0.010)
Woodland:
Decade:
1870- 1880
Quadratic Time Trends by Soil Group
Improved Land,
Land in Farms,
per county acre
per acre in farms
6% - 12%
0% - 6%
6% - 12%
0% - 6%
(6d)
(6a)
(6b)
(6c)
0.016
(0.010)
0.047
(0.035)
[0.45]
0.033**
(0.011)
0.029
(0.025)
[0.29]
0.022*
(0.010)
0.078**
(0.024)
[1.87]
0.039**
(0.009)
0.120**
(0.031)
[4.36]
0.003
(0.012)
0.121**
(0.025)
[3.54]
0.043**
(0.010)
0.097**
(0.027)
[1.92]
0.053**
(0.010)
0.094**
(0.021)
[3.50]
0.024*
(0.010)
0.122**
(0.026)
[2.91]
0.055**
(0.010)
-0.032
(0.026)
[1.94]
1900- 1910
-0.016
(0.011)
[1.98]
0.005
(0.006)
-0.023
(0.024)
[0.44]
-0.012
(0.010)
- 0.038**
(0.011)
[3.36]
-0.001
(0.006)
-0.033
(0.027)
[0.75]
-0.014
(0.010)
- 0.024*
(0.011)
[2.76]
0.006
(0.006)
-0.026
(0.023)
[0.58]
-0.013
(0.009)
1910 - 1920
0.008
(0.011)
[0.58]
0.001
(0.005)
0.034*
(0.017)
[2.17]
-0.004
(0.006)
- 0.019
(0.014)
[1.14]
- 0.004
(0.005)
0.033
(0.021)
[1.63]
0.000
(0.007)
- 0.014
(0.013)
[1.55]
0.005
(0.005)
0.022
(0.016)
[1.26]
0.000
(0.007)
2
R
0.4773
0.5418
Obs.
1885
1885
Notes: This Table presents the full results from Table 4, columns 4 - 6.
0.5095
1885
125
0.5737
1885
0.4864
1885
0.5352
1885
Appendix Table 3. Estimated Changes in the Improvement Intensity of Farmland and Settlement,
Controlling for Initial Outcome Differences
Improved Land,
per acre in farms
Land in Farms,
per county acre
0% - 6%
6% - 12%
0% - 6%
6% - 12%
(7a)
(7b)
(7c)
(7d)
-0.098**
(0.034)
[3.44]
-0.003
(0.013)
- 0.086**
(0.029)
[2.15]
- 0.032**
(0.011)
1880- 1890
0.065*
(0.033)
[2.90]
-0.015
(0.012)
0.031
(0.029)
[1.26]
0.003
(0.013)
1890- 1900
0.076**
(0.021)
[2.95]
0.017
(0.009)
0.062
(0.039)
[0.99]
0.034*
(0.015)
Woodland:
Decade:
1870- 1880
1900 -1910
-0.026*
(0.012)
[2.62]
0.002
(0.006)
-0.045
(0.037)
[0.64]
- 0.026
(0.015)
1910- 1920
-0.005
(0.012)
[0.71]
0.004
(0.005)
0.012
(0.022)
[0.98]
-0.009
(0.010)
R2
0.5735
0.5606
Obs.
1885
1885
Notes: This Table presents the full results from Table 4, column 7.
126
Identifying Agglomeration Spillovers:
Evidence from Million Dollar Plants*
Michael Greenstone
Richard A. Hornbeck
Enrico Moretti
Abstract
The third essay quantifies agglomeration spillovers by comparing the growth of total factor
productivity (TFP) among incumbent plants in "winning" counties that attracted a large
manufacturing plant to "losing" counties that were the plant's second choice. Five years after the
opening, incumbent plants' TFP is 12% higher in winning counties. This effect is larger for
plants with similar labor and technology pools as the new plant. We find evidence of increased
wages in winning counties, indicating that profits increase by less than productivity.
*We thank Daron Acemoglu, Jim Davis, Vernon Henderson, William Kerr, Jeffrey Kling, Jonathan Levin, Stuart
Rosenthal, Christopher Rohlfs, Chad Syverson, and seminar participants at Berkeley, the Brookings Institution,
MIT, NBER Summer Institute, San Francisco Federal Reserve, Stanford, and Syracuse for insightful comments.
Elizabeth Greenwood provided valuable research assistance. The research in this paper was conducted while the
authors were Special Sworn Status researchers of the U.S. Census Bureau at the Boston Census Research Data
Center (BRDC). Support of the Census Research Data Center network from NSF grant no. 0427889 is gratefully
acknowledged. Research results and conclusions expressed are those of the authors and do not necessarily reflect the
views of the Census Bureau. This paper has been screened to insure that no confidential data are revealed.
127
In most countries, economic activity is spatially concentrated. While some of this concentration
is explained by the presence of natural advantages that constrain specific productions to specific
locations, Ellison and Glaeser (1999) and others argue that natural advantages alone cannot
account for the observed degree of agglomeration. Spatial concentration is particularly
remarkable for industries that produce nationally traded goods, because the areas where
economic activity is concentrated are typically characterized by high costs of labor and land.
Since at least Marshall (1890), economists have speculated that this concentration of economic
activity may be explained by cost or productivity advantages enjoyed by firms when they locate
near other firms. The list of potential sources of these agglomerations advantages includes:
cheaper and faster supply of intermediate goods and services; proximity to workers or
consumers; better quality of the worker-firm match in thicker labor markets; lower risk of
unemployment for workers and lower risk of unfilled vacancies for firms following idiosyncratic
shocks; and knowledge spillovers. 1
The possibility of documenting that productivity advantages through agglomeration are
real is tantalizing, because it could provide insights into a series of important questions. Why are
firms that produce nationally traded goods willing to locate in cities like New York, San
Francisco or London, characterized by extraordinary production costs? In general, why do cities
exist and what explains their historical development? Why do income differences persist across
regions and countries?
Beside its obvious interest for urban and growth economists, the existence of
agglomeration spillovers has tremendous practical relevance. Increasingly, local governments
compete by offering substantial subsidies to industrial plants to locate within their jurisdictions.
The main economic rationale for these incentives depends on whether the attraction of new
businesses generates some form of agglomeration externalities. In the absence of positive
externalities, it is difficult to justify the use of taxpayer money for subsidies based on efficiency
grounds. The optimal magnitude of these incentives depends on the magnitude of agglomerations
1 The literature on this topic is enormous, and can not be fully summarized here. Examples include, but are not
limited to, Lucas (1988), Krugman (1991a, 1991b), Henderson (2001a, 2001b, 2003), Davis and Henderson (2004),
Davis and Weinstein (2002), Henderson and Black (1999), Rosenthal and Stange (2001, 2004), Duranton and Puga
(2004), Audretsch and Feldman, (1996, 2004), Moretti (2004a, 2004b, 2004c), Dumais, Ellison and Glaeser (2002),
Glaeser (1999), Ottaviano and Thisse (2004).
128
spillovers, if they exist. 2
Despite their enormous theoretical and practical relevance, the existence and exact
magnitude of agglomeration spillovers are considered open questions by many. To date, there
are two primary approaches for testing for spillovers. The first tests for an unequal distribution
of firms across the country. These "dartboard" style tests reveal that firms are spread unevenly
across the country and that coagglomeration rates are higher between industries that are
economically similar (Ellison, Glaeser and Kerr, 2007). This approach is based on equilibrium
location decisions and does not provide a direct measure of spillovers.
The second approach uses micro data to assess whether firms' total factor productivity
(TFP) is higher when similar firms are located nearby. A notable example is Henderson (2003),
which estimates plant level production functions for machinery and high-tech industries as a
function of the scale of other plants in the same and different industries. 3 The challenge for both
approaches is that firms base their location decisions on where their profits will be highest, and
this could be due to spillovers, natural advantage, or other cost shifters. A causal estimate of the
magnitude of spillovers requires a solution to this problem of identification.
This paper tests for and quantifies agglomeration spillovers by estimating how the
productivity of incumbent manufacturing plants changes when a new, large plant opens in their
county. We estimate augmented Cobb-Douglass production functions that allow TFP to depend
on the presence of the new plant, using plant-level data from the Annual Survey of
Manufacturers. Because the location decision is made to maximize profits, the chosen county is
likely to differ substantially from an average or randomly chosen county both at the time of
opening and in future periods. Valid estimates of the plant opening's spillover effect require the
identification of a county that is identical to the county where the plant decided to locate in the
determinants of incumbent plants' TFP. These determinants are likely to include factors that
affect the new plant's TFP.
This paper's solution is to rely on the reported location rankings of profit-maximizing
firms to identify a valid counterfactual for what would have happened to incumbent plants' TFP
in winning counties in the absence of the plant opening.
2 We
These rankings come from the
discuss in more detail the policy implications of local subsidies in Greenstone and Moretti (2004). See also
Card, Hallock, and Moretti (2007) and Glaeser (2001).
3 Moretti (2004b) takes a similar approach to estimate agglomeration externalities generated by human capital
spillovers.
129
corporate real estate journal Site Selection, which includes a regular feature titled "Million Dollar
Plants" that describes how a large plant decided where to locate. When firms are considering
where to open a large plant, they typically begin by considering dozens of possible locations.
They subsequently narrow the list to roughly 10 sites, among which 2 or 3 finalists are selected.
The "Million Dollar Plants" articles report the county that the plant ultimately chose (i.e., the
'winner'), as well as the one or two runner-up counties (i.e., the 'losers'). The losers are counties
that have survived a long selection process, but narrowly lost the competition.
The identifying assumption is that the incumbent plants in the losing county form a valid
counterfactual for the incumbents in the winning counties, after conditioning on differences in
pre-existing trends, plant fixed effects, industry-by-year fixed effects, and other control
variables. Compared to the rest of the country, winning counties have higher rates of growth in
income, population, and labor force participation. But compared to losing counties in the years
before the opening of the new plant, winning counties have similar trends in most economic
variables. This finding is consistent with both our presumption that the average county is not a
credible counterfactual and our identifying assumption that the losers form a valid counterfactual
for the winners.
We first measure the effect of the new Million Dollar Plant (MDP) on total factor
productivity (TFP) of all incumbent manufacturing plants in winning counties. In the 7 years
before the MDP opened, we find statistically equivalent trends in TFP for incumbent plants in
winning and losing counties, which supports the validity of the identifying assumption. After the
MDP opened, incumbent plants in winning counties experienced a sharp relative increase in
TFP. Five years later, the MDP opening is associated with a 12% relative increase in incumbent
plants' TFP. 4
This effect is statistically significant and economically substantial; on average
incumbent plants' output in winning counties is $430 million higher 5 years later (relative to
incumbents in losing counties), holding constant inputs. We interpret this finding as evidence of
meaningful productivity spillovers generated by increased agglomeration.
Having found evidence in favor of the existence of agglomeration spillovers, we then try
to shed some light on the possible mechanisms. We follow Moretti (2004b) and Ellison, Glaeser,
and Kerr (2007) and investigate whether the magnitude of the spillovers depends on economic
4 Notably,
naive estimates that control for observables but do not use the MDP research design find negative
productivity effects.
130
linkages between the incumbent plant and the MDP. Specifically, we test whether incumbents
that are geographically and economically linked to the MDP experience larger spillovers, relative
to incumbents that are geographically close but economically distant from the MDP. We use
several measures of economic links including input and output flows, measures of the degree of
sharing of labor pools, and measures of technological linkages.5
We find that spillovers are larger for incumbent plants in industries that share worker
flows with the MDP industry. A one standard deviation increase in our measure of worker
transitions is associated with a 7 percentage point increase in the magnitude of the spillover.
Similarly, the measures of technological linkages indicate statistically meaningful increases in
the spillover effect. Surprisingly, we find little support for the importance of input and output
flows in determining the magnitude of the spillover. Overall, this evidence provides some
support for the notion that spillovers occur between firms that share workers and use similar
technologies.
To guide the analysis and interpret the results, we set out a straightforward Roback
(1982) style model that incorporates spillovers between producers and derives an equilibrium
allocation of firms and workers across locations. In the model, the entry of a new firm produces
spillovers. This leads new firms who are interested in gaining access to the spillover to enter.
This process of entry leads to competition for inputs so that incumbent firms face higher prices
for labor, land, and other local inputs. In the model, firms produce nationally traded goods, so
they cannot raise the price of their output in response to the higher input prices. Thus, the long
run equilibrium is obtained when the value of the increase in output due to spillovers is equal to
the increased costs of production due to the higher input prices
The data support two predictions derived from this model. First, we find positive net
entry in winning counties, which the model predicts will occur if there are sufficiently large
positive spillovers.
Second, we find increases in quality-adjusted wages following MDP
openings. These higher wages are consistent with the documented increase in economic activity
in the winning counties (which presumably is a response to the spillovers). Furthermore, the
higher wages support the model's prediction that the productivity gains from agglomeration do
not necessarily translate into higher profits for incumbents in the long run.
5 We
are deeply indebted to Glenn Ellison, Edward Glaeser, and William Kerr for providing their data for five of
these measures of economic distance.
131
The remainder of the paper is organized as follows. Section I presents a simple model.
Section II discusses the identification strategy. Section III presents the data sources. Section IV
presents the econometric model. Sections V and VI presents the empirical findings and interpret
them. Section VII concludes.
I. Theories of Agglomeration and Theoretical Framework
We are interested in identifying how the opening of a new plant in a county affects the
productivity, profits, and input use of existing plants in the same county. We begin by reviewing
the theories of agglomeration. We then present a simple theoretical framework that guides the
subsequent empirical exercise and aids in interpreting the results.
A. Theories of Agglomeration
Economic activity is geographically concentrated (Ellison and Glaeser, 1997). What are
the forces that can explain such agglomeration of economic activity? Here we summarize five
possible reasons for agglomeration, and briefly discuss what each of them implies for the
relationship between productivity and the density of economic activity.
(1) First, it is possible that firms (and workers) are attracted to areas with a high
concentration of other firms (and other workers) by the size of the labor market. There are at
least two different reasons why larger labor markets may be attractive. First, a thick labor
market is beneficial in the presence of search frictions, if jobs and workers are heterogeneous. In
the presence of frictions, a worker-firm match will be on average more productive in areas where
there are many firms offering jobs and many workers looking for jobs.6
Alternatively, it is possible that large labor markets are more desirable because they
provide insurance against idiosyncratic shocks, either on the firm side or on the worker side
(Krugman 1991a). If moving is costly for workers and firms are subject to idiosyncratic and
unpredictable demand shocks that lead to lay-offs, workers will prefer to be in areas with thick
labor markets to reduce the probability of being unemployed. Similarly, if finding new workers
is costly, firms will prefer to be in areas with thick labor markets to reduce the probability of
6
For a related point in a different context, see Petrongolo and Pissarides (2005).
132
having unfilled vacancies. 7
These two hypotheses have different implications for the relationship between
concentration of economic activity and productivity. If the size of the labor market leads only to
better worker-firm matches, we should see that firms located in denser areas are more productive
than otherwise identical firms located in less dense areas. The exact form of this productivity
gain depends on the shape of the production function. 8
On the other hand, if the only effect of thickness in labor market is a lower risk of
unemployment for workers and a lower risk of unfilled vacancies for firms, there should not be
differences in productivity between dense and less dense areas. While productivity would not
vary, wages would vary across areas depending on the thickness of the labor market, although
the exact effect of density on wages is a priori ambiguous. 9 This change in relative factor prices
will change the relative amount of labor and capital used. Unlike the case of improved matching
described above, the production function does not change: for the same set of labor and capital
inputs, the output of firms in denser areas should be similar to the output of firms in less dense
areas.
(2) A second reason why the concentration of economic activity may be beneficial has to
do with transportation costs (Krugman 1991a and 1991b, Glaeser and Kohlhase, 2003). Because
in this paper we focus on firms that produce nationally traded goods, transportation costs of
finished products are unlikely to be the relevant cost in this paper's setting. Only a small fraction
of buyers of the final product is likely to be located in the same area as our manufacturing plants.
The relevant costs are the transportation costs of suppliers of local services and local
intermediate goods. Firms located in denser areas are likely to enjoy cheaper and faster delivery
7A
third alternative hypothesis has to do with spillovers that arise because of endogenous capital accumulation. For
example, in Acemoglu (1996), plants have more capital and better technology in areas where the number of skilled
workers is larger. If firms and workers find each other via random matching and breaking the match is costly,
externalities will arise naturally even without learning or technological externalities. The intuition is simple. The
privately optimal amount of skills depends on the amount of physical capital a worker expects to use. The privately
optimal amount of physical capital depends on the number of skilled workers. If the number of skilled workers in a
city increases, firms in that city, expecting to employ these workers, will invest more. Because search is costly,
some of the workers end up working with more physical capital and earn more than similar workers in other cities.
8 For example, it is possible that the productivities of both capital and labor benefit from the improved match in
denser areas. It is also possible that the improved match caused by a larger labor market benefits only labor
productivity. This has different implications for the relative use of labor and capital, but total factor productivity
will be higher regardless.
9 Its sign depends on the relative magnitude of the compensating differential that workers are willing to pay for
lower risk of unemployment (generated by an increase in labor supply in denser areas) and the cost savings that
firms experience due to lower risk of unfilled vacancies (generated by an increase in labor demand in denser areas).
133
of local services and local intermediate goods. For example, a high-tech firm that needs a
specialized technician to fix a machine is likely to get service more quickly and at lower cost if it
is located in Silicon Valley than in the Nevada desert.
This type of agglomeration spillover does not imply that the production function varies as
a function of density of economic activity: for the same set of labor and capital inputs, the output
of firms in denser areas should be similar to the output of firms in less dense areas. However,
production costs should be lower in denser areas.
(3) A third reason why the concentration of economic activity may be beneficial has to do
with knowledge spillovers. There are at least two different versions of this hypothesis. First,
economists and urban planners have long speculated that the sharing of knowledge and skills
through formal and informal interaction may generate positive production externalities across
workers. 10 Empirical evidence indicates that this type of spillover may be important in some
high-tech industries. For example, patent citations are more likely to come from the same state
or metropolitan area as the originating patent (Jaffe et al. 1993). Saxenian (1994) argues that
geographical proximity of high-tech firms in Silicon Valley is associated with a more efficient
flow of new ideas and ultimately causes faster innovation." Second, it is also possible that
proximity results in sharing of information on new technologies and therefore leads to faster
technology adoption. This type of social learning phenomenon applied to technology adoption
was first proposed by Griliches (1958).
If density of economic activity results in intellectual externalities, the implication of this
type of agglomeration model is higher productivity. In particular, we should see that firms
located in denser areas are more productive than otherwise identical firms located in less dense
areas. As with the search model, this higher productivity could benefit both labor and capital, or
only one of the two factors, depending on the form of the production function. On the other
hand, if density of economic activity only results in faster technology adoption and the price of
new technologies reflects their higher productivity, there should be no relationship between
productivity and density, after properly controlling for quality of capital.
(4) It is possible that firms concentrate spatially not because of any technological
10 See, for example, Marshall (1890), Lucas (1988), Jovanovic and Rob (1989), Grossman and Helpman (1991),
Saxenian (1994), Glaeser (1999), and Moretti (2004a, 2004b and 2004c).
" The entry decisions of new biotechnology firms in a city depend on the stock of outstanding scientists there, as
measured by the number of relevant academic publications (Zucker et al., 1998). Moretti (2004b) finds stronger
human capital spillovers between pairs of firms in the same city that are economically or technologically closer.
134
spillover, but because local amenities valued by workers are concentrated. For example, skilled
workers may prefer certain amenities more than unskilled workers. This would lead firms that
employ relatively more skilled workers to concentrate in locations where these amenities are
available. In this case, we should not see any difference in productivity between dense areas and
less dense areas, although we should see differences in wages that reflect the compensating
differential.
(5) Finally, spatial concentration of some industries may be explained by the presence of
natural advantages. For example, the oil industry is concentrated in a limited number of states
because those states have the most accessible oil fields. Similarly, the wine industry is
concentrated in California because that is where good weather and suitable land are to be found.
For some manufacturing productions, the presence of a harbor may be important. The presence
of natural advantages has the implication that firms located in areas with high concentration of
similar firms are more productive, but of course this correlation has nothing to do with
agglomeration spillovers. Since most natural advantages are fixed over time, this explanation is
not particularly relevant for our empirical estimates, which exploit variation over time in
agglomeration.
B. A Simple Model
We begin by considering the case where incumbent firms are homogenous in size and
technology.
Later we consider what happens when incumbent firms are heterogeneous.
Throughout the paper, we focus on the case of factor-neutral spillovers.
(a) Homogeneous Incumbents. We assume that all incumbent firms use a production
technology that uses labor, capital, and land to produce a nationally traded good whose price is
fixed and is normalized to 1. Incumbent firms choose their amount of labor, L, capital, K, and
land, T, to maximize the following expression:
Max L,K,T { f[A, L, K, T] - w L - r K - q T }
where w, r and q are input prices and A is a productivity shifter (TFP). Specifically, A includes
all factors that affect the productivity of labor, capital, and land equally, such as technology and
agglomeration spillovers, if they exist.
In particular, to explicitly allow for agglomeration
effects, we allow A to depend on the density of economic activity in an area:
(1)
A = A(N)
135
where N is the number of firms that are active in a county, and all counties have equal size. We
define factor-neutral agglomeration spillovers as the case where A increases in N:
6A /6N >0
If instead (6A /6N) =0, we say that there are no factor-neutral agglomeration spillovers.
Let L*(w,r,q) be the optimal level of labor inputs, given the prevailing wage, cost of
capital, and cost of industrial land. Similarly, let K*(w,r,q) and T*(w,r,q) be the optimal level of
capital and land, respectively. In equilibrium, L*, K*, and T* are set so that the marginal product
of each of the three factors is equal to its price.
We assume that capital is internationally traded, so its price does not depend on local
demand or supply conditions. However, we allow for the price of labor and land to depend on
local economic conditions. In particular, we allow the supply of labor and land to be less than
infinitely elastic at the county level.
We attribute the upward sloping labor supply curve to the existence of moving costs.
Like in the standard Roback (1982) model, we assume that workers' indirect utility depends on
wages and cost of housing, and that in equilibrium workers are indifferent across locations.
Workers are mobile across locations, but unlike the standard Roback (1982) model we allow for
moving costs. For simplicity, we ignore labor supply decisions within a given location and
assume that all residents provide a fixed amount of labor.
To illustrate this, consider that there are m workers in county c before the opening of the
new plant. In particular, m is such that, given the distribution of wages and the housing costs
across localities, the marginal worker in another county is indifferent between moving to county
c and staying. When a new plant opens in county c, wages there start rising, and some workers
find it optimal to move to c. The number of workers who move, and therefore the slope of the
labor supply function, depend on the shape of the mobility cost function. Let w(N) be the
inverse of the reduced-form labor supply function that links the number of firms, N, active in a
county to the local nominal wage level, w.
Similarly, we allow the supply of industrial land to be less than infinitely elastic at the
county level. For example, it is possible that the supply of land is fixed because of geography or
land-use regulations. Alternatively, it may not be completely fixed, but it is possible that the best
industrial land has already been developed, so that the marginal land is of decreasing quality or
more expensive to develop. Irrespective of the reason, we call q(N) the (inverse of the) reduced
136
form land supply function that links the number of firms, N, to the price of land, q. We can
therefore write the equilibrium level of profits, I-*, as
11* = f[ A(N), L*(w(N), r, q(N)), K*(w(N), r, q(N)), T*(w(N), r, q(N))] - w(N) L*(w(N), r, q(N)) - r K*(w(N), r, q(N)) - q(N) T*(w(N), r, q(N))
where we now make explicit the fact that TFP, wages, and land prices depend on the number of
firms active in a county.
Consider the total derivative of incumbents' profits with respect to a change in the
number of firms:
(2)
dlI*/dN = (6f /6A 6A/6N)
+ 6w/6N { [6L*/6w (6f/6L - w) - L*] + [6K*/6w (6f/8K - r)] + [6T*/6w (6f/6T -
q)] }
+ 6q/6N { [6L*/Sq (6f/6L - w)] + [6K*/5q (6f/6K - r)] + [6T*/aq (6f/6T - q) - T*]
}
If all firms are price takers and all factors are paid their marginal product, equation (2) simplifies
considerably and can be written as:
dFI*/dN = (6f/SA 6A/5N) - [ 6w/6N L* + 6q/6N T*].
(3)
Equation (3) makes clear that the effect of an increase in N is the sum of two opposite
effects. First, if there are positive spillovers, the productivity of all factors increases. In equation
(3), this effect on TFP is represented by the first term, (6f /6A 6A/6N).
This effect is
unambiguously positive, because it allows an incumbent firm to produce more output using the
same amount of inputs. Formally, 6f /6A >0 by assumption and, if there are positive spillovers,
6A/6N >0.
The second term, - [ 6w/6N
L* + 6q/6N
T*], represents the negative effect from
increases in the cost of production, specifically the prices of labor and land. Formally, this term
is negative because we have assumed that 6w/6N > 0 and 6q/6N > 0. (The magnitudes depend
on the elasticity of the supply of labor and land.) Intuitively, an increase in N is an increase in
the level of economic activity in the county and therefore an increase in the local demand for
labor and land.
Unlike the beneficial effect of agglomeration spillovers, the increase in factor prices is
costly for incumbent firms, because they now have to compete for locally scare resources with
the new entrant. The increase in wages and land prices has two effects on incumbents. First, for
137
a given level of input utilizations, it mechanically raises production costs. Second, it leads the
firm to re-optimize and to change its use of the different production inputs. In particular, given
that the price of capital is not affected by an increase in N, the firm is likely to end up using more
capital than before:
6K*/6N =>0.
By contrast, the effect on the use of labor and land is ambiguous. On the one hand, the
productivity of all factors increases. On the other hand, the price of labor and land might
increase. The net effect depends on the magnitude of the factor price increases, as well as on the
exact shape of the production function (i.e., the strength of technological complementarities
between labor, capital, and land).
It is instructive to apply these derivations to the case of a MDP opening that causes
positive spillovers. We initially consider the case where for incumbent firms dl*/dN < 0. This
would occur when the agglomeration spillover is smaller than the increase in production costs.
In this case, the MDP's opening would not lead to any entry and could cause some existing firms
to exit. 12
The alternative case is that dFI*/dN > 0, which occurs when the magnitude of the spillover
due to the MDP's opening exceeds the increase in factor prices due to the MDP's demand for
local inputs. In the short run, profits will be positive for new entrants. These positive profits
will disappear over time as the price of local factors, like land and possibly labor, is bid up.
In the long run, there is an equilibrium such that firms and workers are indifferent
between the county where the new plant has opened and other locales. Since the amount of land
is fixed, the higher levels of productivity are likely to be capitalized into land prices. It is also
likely that wages will increase. This may occur due to moving costs as noted above. 13 These
adjustments make workers indifferent between the county with the new plant and other counties.
Similarly, the changes in factor prices mean that firms earn the same profits in the county with
the new plant (even in the presence of the spillovers) and in other locations. From a practical
perspective, it is impossible to know when the short run ends and the long run begins.
There are two empirical predictions that apply when there are positive spillovers. First, if
12 Similarly,
if the spillovers are zero or negative, one might expect exit of incumbent firms and a reduction in local
economic activity.
13 Even with zero moving costs and an infinitely elastic supply of labor,
wages will increase if there are land price
increases as workers will demand higher wages as compensation for the higher land rents for their homes (Roback
1982).
138
the magnitude of the spillovers is large enough, new firms will enter the MDP's county to gain
access to the spillover. This prediction of increased economic activity holds at any point after
potential new entrants have had sufficient time to respond. The second prediction is that the
prices of locally traded inputs will rise as the MDP and the new entrants bid for these inputs. 14
(b) Heterogeneous Incumbents. What happens if the population of incumbent firms is nonhomogeneous? Consider the case where there are two types of firms: high-tech and low-tech.
Assume that for technological reasons, the type of workers employed by high-tech firms, LH,
differs to some extent from the type of workers employed by low-tech firms, LL, although there
is some overlap.
Assume that the new entrant is a high-tech firm.
Equations (4) and (5)
characterize the effect of the new high-tech firm on high-tech and low-tech incumbents:
6
= (6fH/ AH
6AH/6NH) - [ 6WH/ 6 NH L*H + 6q/6NH T*]
(4)
dlH*/dNH
(5)
dflL*/dNH = (6fL/6AL 6AL/6NH) - [6wL/6NH L*L + 6q/6NH T*]
It is plausible to expect that the beneficial effect of agglomeration spillovers generated by a
new high-tech entrant is larger for high-tech firms than for low-tech firms:
(5')
(6f H/6 AH 6AH/6NH) >(6fL/ 6 AL 6AL/6NH)
At the same time, one might expect that the increase in labor costs is also higher for the hightech incumbents, given that they are now competing for workers with an additional high-tech
firm:
6
WH/ 6 NH > 6 wL/ 6 NH
The effect on land prices should be similar for both firm types, since the assumption of a single
land market seems reasonable.
There are two takeaways here. First, it may be reasonable to expect larger spillovers to firms
This paper focuses on the case where the productivity benefits of the agglomeration spillovers are distributed
equally across all factors. What happens when agglomeration spillovers are factor biased? Assume, for example,
that agglomeration spillovers raise the productivity of labor, but not the productivity of capital. Like before, the
technology is f[A, L, K, T], but now L represents units of effective labor. In particular, L = OH, where H is the
number of physical workers and 0 is a productivity shifter. We define factor-biased agglomeration spillover as the
case where the productivity shifter 0 depends positively on the density of the economic activity in the county 0 =
0(N) and 60/6N >0. If 6A /6N =0 and factors are paid their marginal product, then the effect of an increase in the
density of the economic activity in a county on incumbent firms simplifies to dfl*/dN = (6f /6H 60/5N) H - [ 6w/6N
14
H* + 6q/6N T*]. The effect on profits can be decomposed in two parts. The first term represents the increased
productivity of labor. It is the product of the sensitivity of output to labor (6f /6H >0), times the magnitude of the
agglomeration spillover (60/6N > 0 by definition), times the number of workers. The second term is the same as in
equation (3), and represents the increase in the costs of locally supplied inputs. The increase in N changes the
optimal use of the production inputs. Labor is now more productive, and its equilibrium use increases: 6L*/6N <=0.
Land is equally productive but its price increases. Its equilibrium use declines: 6T*/6N <=0. Neither the price nor
the productivity of capital is affected by an increase in N. Its equilibrium use depends on technology. Specifically,
it depends on the elasticity of substitution between labor and capital.
139
that are economically "close" to the new plant. Second the relative impact of the new plant on
profits is unclear, because the economically "closer" plants are likely to have bigger spillovers
and larger increases in production costs.
C. Empirical Predictions
The simple theoretical framework above generates four predictions that we bring to the
data. Specifically if there are positive spillovers, then:
1. the opening of a new plant will increase the TFP of incumbent plants.
2. the increase in TFP may be larger for firms that are economically "closer" to the new
plant.
3. the density of economic activity in the county will increase as firms move in to gain
access to the positive spillovers (if the spillovers are large enough).
4. the price of factors of production that are traded locally will increase. We test for
changes in the price of quality-adjusted labor, which is arguably the most important
locally supplied factor of production for manufacturing establishments.
II. Plant Location Decisions and Research Design
In testing the four empirical predictions outlined above, the main econometric challenge
is the fact that firms do not choose their location randomly. Firms are profit maximizers and
choose to locate where their expectation of the present discounted value of the stream of future
profits is greatest. This net present value varies tremendously across locations, depending on
many factors, including transportation infrastructure, the availability of workers with particular
skills, subsidies, etc. These factors are frequently unobserved. Further, they are likely to be
correlated with the TFP of existing plants.
Therefore, a naive comparison of the TFP of incumbents in counties that experience a
plant opening with the TFP of incumbents in counties that do not experience a plant opening is
likely to yield biased estimates of productivity spillovers. Credible estimates of the impact of a
plant opening on TFP of incumbent plants require the identification of a location that is similar to
the location where the plant decided to locate in the determinants of incumbent plants' TFP.
This section provides a case study for how BMW picked the location for one of its
140
plants. 15 The intent is to demonstrate the empirical difficulties that arise when estimating the
effect of plant openings on the TFP of incumbent plants. Further, it illustrates informally how
our research design may circumvent these difficulties.
After overseeing a worldwide competition and considering 250 potential sites for its new
plant, BMW announced in 1991 that they had narrowed the list of potential candidates to 20
counties.
Six months later, BMW announced that the two finalists in the competition were
Greenville-Spartanburg, South Carolina, and Omaha, Nebraska. In 1992, BMW announced that
they would site the plant in Greenville-Spartanburg and that they would receive a package of
incentives worth approximately $115 million funded by the state and local governments.
Why did BMW choose Greenville-Spartanburg? Two factors were important in this
decision. The first was BMW's expected future costs of production in Greenville-Spartanburg,
which are presumably a function of the county's expected supply of inputs and BMW's
production technology.
According to BMW, the characteristics that made Greenville-
Spartanburg more attractive than the other 250 sites initially considered were: low union density,
a supply of qualified workers, the numerous global firms, including 58 German companies, in
the area; the high quality transportation infrastructure, including air, rail, highway, and port
access; and access to key local services.
For our purposes, the important point to note here is that these county characteristics are a
potential source of unobserved heterogeneity. While these characteristics are well documented
in the BMW case, they are generally unknown and unobserved. If these characteristics also
affect the growth of TFP of existing plants, a standard regression that compares GreenvilleSpartanburg with the other 3,000 United States counties will yield biased estimates of the effect
of the plant opening. A standard regression will overestimate the effect of plant openings on
outcomes if, for example, counties that have more attractive characteristics (e.g., improving
transportation infrastructure) tend to have faster TFP growth. Conversely, a standard regression
would underestimate the effect if, for example, incumbent plants' declining TFP encouraged new
entrants (e.g., cheaper availability of local inputs).
A second important factor in BMW's decision was the value of the subsidy it received.
Presumably Greenville-Spartanburg was willing to provide BMW with $115 million in subsidies
'~ This plant is in Greenstone and Moretti's (2004) set of 82 MDP plants. Due to Census confidentiality restrictions,
we cannot report whether this plant is part of this paper's analysis.
141
because it expected economic benefits from BMW presence. According to local officials, the
facility's ex-ante expected five-year economic impact on the region was $2 billion. As a part of
this $2 billion, the plant was expected to create 2,000 jobs directly and another 2,000 jobs
indirectly. In principle, these 2,000 additional jobs could reflect the entry of new plants or the
expansion of existing plants caused by agglomeration economies. (The empirical section tests
whether this is indeed the case on average). Thus, the subsidy is likely to be a function of the
expected gains from agglomeration for the county.'16
This possibility is relevant for this paper's identification strategy, because the magnitude
of the spillover from a particular plant depends on the level and growth of a county's industrial
structure, labor force, and a series of other unobserved variables. For this reason, the factors that
determine the total size of the potential spillover (and presumably the size of the subsidy)
represent a second potential source of unobserved heterogeneity.
If this unobserved
heterogeneity is correlated with incumbent plants' TFP, standard regression equations will be
misspecified due to omitted variables, just as described above.
In order to make valid inferences in the presence of the heterogeneity associated with the
plant's expected local production costs and the county's value of attracting the plant, knowledge
of the exact form of the selection rule that determines plants' location decisions is generally
necessary. As the BMW example demonstrates, the two factors that determine plant location
decisions are generally unknown to researchers and, in the rare cases where they are known, are
difficult to measure. Thus, the effect of a plant opening on incumbents' TFP is very likely to be
confounded by differences in factors that determine the plants' profitability at the chosen
location.
As a solution to this identification problem, we rely on the reported location rankings of
profit-maximizing firms to identify a valid counterfactual for what would have happened to
incumbent plants in winning counties in the absence of the plant opening. We implement the
research design using data from the corporate real estate journal Site Selection. Each issue of this
journal includes an article titled the "Million Dollar Plants" that describes how a large plant
decided where to locate. These articles always report the county that the plant chose (i.e., the
16
The fact that business organizations such as the Chambers of Commerce support these incentive plans (as was
the
case with BMW) suggests that incumbent firms expect such increases. Greenstone and Moretti (2004) present a
model that describes the factors that determine local governments' bids for these plants and whether successfully
attracting a plant will be welfare increasing or decreasing for the county.
142
'winner'), and usually report the runner-up county or counties (i.e., the "losers").
17
As the BMW
case study indicates, the winner and losers are usually chosen from an initial sample of "semifinalist" sites that in many cases number more than a hundred. 18 The losers are counties that have
survived a long selection process, but narrowly lost the competition.
We use the losers to identify what would have happened to the productivity of incumbent
plants in the winning county in the absence of the plant opening. Specifically, we assume that
incumbent firms' TFP would have trended identically in the absence of the plant opening in
winning and losing counties within a case.
In practice, we adjust for covariates so our
identifying assumption is weaker. The subsequent analysis provides evidence that supports the
validity of this assumption. Even if this assumption fail to hold, we presume that this pairwise
approach is more reliable than using regression adjustment to compare the TFP of incumbent
plants in counties with new plants to the other 3,000 United States counties or to using a
matching procedure based on observable variables. 19
III. Data Sources and Summary Statistics
A. Data Sources
The "Million Dollar Plants" articles typically reveal the county where the new firm (the
"Million Dollar Plant") ultimately chose to locate (the "winning county"), as well as the one or
two runner-up counties (the "losing counties"). The articles tend to focus on large plants that are
the target of local government subsidies. Important limitations of these articles are that the
magnitude of the subsidy offered by the winning counties is in many cases unobserved and that
the bid is almost always unobserved for losing counties.
We identify the Million Dollar Plants in the Standard Statistical Establishment List
(SSEL) - which is the Census Bureau's "most complete, current, and consistent data for U.S.
17 In
some instances the "Million Dollar Plants" articles do not identify the runner-up county. For these cases, we
did a Lexis/Nexis search for other articles discussing the plant opening and in 4 cases, among the original 82, we
were able to identify the losing counties. The Lexis/Nexis searches were also used to identify the plant's industry
when this was unavailable in Site Selection. Comprehensive data on the subsidy offered by winning and losing
counties is unavailable in the Site Selection articles.
18 The names of the semi-finalists are rarely reported.
19Propensity score matching is an alternative approach (Rosenbaum and Rubin 1983). Its principal shortcoming
relative to our approach is its assumption that the treatment (i.e., winner status) is "ignorable" conditional on the
observables. As it should be clear from the example, adjustment for observable variables through the propensity
score is unlikely to be sufficient.
143
business establishment" 20 - and matched the plants to the Annual Survey of Manufactures (ASM)
and the Census of Manufactures (CM) from 1973-1998.21
Of the 82 MDP openings in
Greenstone and Moretti (2004), we identified 47 genuine and useable MDP openings in the
manufacturing data. In order to qualify as a genuine and useable MDP manufacturing opening,
we imposed the following criterion: 1) there had to be a new plant in the manufacturing sector,
owned by the reported firm, appearing in the SSEL within 2 years before and 3 years after the
publication of the MDP article; 2) the plant identified in the SSEL had to be located in the county
indicated in the MDP article; and 3) there had to be incumbent plants in both winning and losing
counties that were there for each of the previous 8 years. Among the 35 MDP openings that did
not qualify, roughly 20 were outside of the manufacturing sector. (We cannot report the exact
number because of the Census Bureau's confidentiality rules).
To obtain information on incumbent establishments in winner and loser counties, we use
the ASM and CM. The ASM and CM contain information on employment, capital stock, total
value of shipments, plant age, and firm identifiers. The 4-digit SIC code and county of location
are also reported and these play a key role in the analysis. Importantly, the manufacturing data
contain a unique plant identifier, making it possible to follow individual plants over time.22 The
sample that we use includes plants that were continuously present in the ASM in the 8 years
preceding the year of the plant opening plus the year of the opening. Additionally, we drop
observations on plants that have the same owner as the MDP plants. In this period, the ASM
sampling scheme was positively related to firm and plant size. Any establishment that was part
of a company with manufacturing shipments exceeding $500 million was sampled with certainty,
as were establishments with 250 or more employees.
There are a few noteworthy features of this sample of potentially affected plants. First,
the focus on existing plants allows for a test of spillovers on a fixed sample of pre-existing
plants, which eliminates concerns related to the endogenous opening of new plants and
compositional bias. Second, it is possible to form a genuine panel of manufacturing plants.
Third, a disadvantage is that the results may not be externally valid to smaller incumbent plants
The SSEL is confidential and was accessed in a Census Data Research Center. The SSEL is updated continuously
and incorporates data from all Census Bureau economic and agriculture censuses and current business surveys,
quarterly and annual Federal income and payroll tax records, and other Departmental and Federal statistics and
administrative records programs.
20
21 The sample is cut at 1998 because sampling methods in the ASM changed for 1999. The sample begins in 1973
because of minor known inconsistencies with the 1972 CM.
22 See the appendix in Davis, Haltiwanger, and Schuh (1996) for a more thorough
description of the ASMand CM.
144
that are not sampled with certainty throughout this period. Nevertheless, it is relevant that this
sample of plants accounts for 54% of county-wide manufacturing shipments in the last CM
before the MDP opening.
Besides testing for an average spillover effect, we also test whether the estimated
agglomeration effects are larger in industries that are more closely linked to the MDP based on
some measure of economic distance.
We use six measures of economic distance in three
categories. First, to measure supplier and customer linkages, we use data on the fraction of each
industry's manufactured inputs that come from each 3-digit industry and the fraction of each
industry's outputs sold to manufacturers that are purchased by each 3-digit industry. Second, to
measure the frequency of worker mobility between industries, we use data on labor market
transitions from the Current PopulationSurvey (CPS) outgoing rotation file. In particular, we
measure the fraction of separating workers from each 2-digit industry that move to firms in each
2-digit industry. Third, to measure technological proximity, we use data on the fraction of
patents manufactured in a 3-digit industry that cite patents manufactured in each 3-digit industry.
We also use data on the amount of R&D expenditure in a 3-digit industry that is used in other 3digit industries.
Finally, one further data issue merits attention. We have two sources of information on
the date of the plant opening. The first is the MDP articles, which often are written when ground
is broken on the plant but other times are written when the location decision is made or the plant
begins operations. The second source is the SSEL, which in principle reports the plant's first
year of operation. However, it is known that plants occasionally enter the SSEL after their
opening.
Thus, there is uncertainty about the date of the plant's opening.
Further, the date at
which the plant could affect the operations of existing plants depends on the channel for any
agglomeration economies.
If the agglomeration economies are a consequence of supplier
relationships, then they could occur as soon as the plant is announced. For example, the new
plant's management might visit existing plants and provide suggestions on operations.
Alternatively, the agglomeration spillovers may be driven by the labor market and therefore may
depend on sharing labor. In this case, agglomeration economies may not be evident until the
plant is operating. Based on these data and conceptual issues, there is not clear guidance on
when the new plant could affect other plants. Rather than take an unsupportable stand, we
145
emphasize results using the earlier of the year of the publication of the magazine article and the
year that the new plant appears in the SSEL. We also report separate results based on using
these two years as the opening date. The basic findings are robust to these alternatives.
B. Summary Statistics
Table 1 presents summary statistics on the sample of plant location decisions that forms
the basis of the analysis. As discussed in the previous subsection, there are 47 manufacturing
MDP openings that we can match to plant level data. There are plants in the same 2-digit SIC
industry in both winning and losing counties in the 8 years preceding the opening for just 16 of
these openings.
The table reveals some other facts about the plant openings. 23 We refer to the winner and
accompanying loser(s) associated with each plant opening announcement as a "case." There are
two or more losers in 16 of the cases, so there are a total of 73 losing counties along with 47
winning counties. Some counties appear multiple times in the sample (as either a winner or
loser), and the average county in the sample appears a total of 1.09 times. The difference
between the year of the MDP article's publication and the year the plant appears in the SSEL is
roughly spread evenly across the categories -2 to -1 years, 0 years, and 1 to 3 years. For clarity,
positive differences refer to cases where the article appears after the plant is identified in the
SSEL. The date of the plant openings ranges from the early 1980s through the early 1990s.
The remainder of Table 1 provides summary statistics on the new MDP plants five years
after their assigned opening date to provide a sense of their magnitude. These MDPs are quite
large: they are more than twice the size of the average incumbent plant and account for roughly
nine percent of the average county's total output one year prior to its opening.
Table 2 provides summary statistics on the measures of industry linkages and further
descriptions of these variables. In all cases, the proximity between industries is increasing in the
value of the variable. For ease of interpretation in the subsequent regressions, these variables are
normalized to have a mean of zero and a standard deviation of one.
Table 3 presents the means of county-level and plant-level variables across counties.
These means are reported for winners, losers, and the entire United States in columns (1), (2),
A number of the statistics in Table 1 are reported in broad categories to comply with the Census Bureau's
confidentiality restrictions and to avoid disclosing the identities of any individual plants.
23
146
and (3), respectively. 24 In the winner and loser columns, the plant-level variables are calculated
among the incumbent plants present in the ASM in the 8 years preceding the assigned opening
date and the assigned opening date. All entries in the entire United States column are weighted
across years to produce statistics for the year of the average MDP opening in our sample.
Further, the plant characteristics are only calculated among plants that appear in the ASM for at
least 9 consecutive years. Column (4) presents the t-statistics from a test that the entries in (1)
and (2) are equal, while Column (5) repeats this for a test of equality between columns (1) and
(3). Columns (6) through (10) repeat this exercise among the cases where there are plants within
the same 2-digit SIC industry as the MDP plant. In these columns, the plant characteristics are
calculated among the plants in the same 2-digit industry.
This exercise provides an opportunity to assess the validity of the research design, as
measured by pre-existing observable county and plant characteristics. To the extent that these
observable characteristics are balanced among winning and losing counties, this should lend
credibility to the analysis. The comparison between winner counties and the rest of the United
States provides an opportunity to assess the validity of the type of analysis that would be
undertaken in the absence of a quasi-experiment.
The top panel reports on county-level characteristics measured in the year before the
assigned plant opening and the percentage change between 7 years and 1 year before the
opening.
It is evident that compared to the rest of the country, winning counties have higher
incomes, population and population growth, labor force participation rates and growth, and a
higher share of labor in manufacturing.
Among the 8 variables in this panel, 6 of the 8
differences would be judged to be statistically significant at conventional levels.
These
differences are substantially mitigated when the winners are compared to losers, and this is
reflected in the fact that 3 of the 8 variables are statistically different at the 5% level but none are
at the 1% level. Notably, the raw differences between winners and losers within the subset of
cases where there are plants in the same 2 digit SIC industry are generally smaller, and none of
them would be judged to be statistically significant.
The second panel reports on the number of sample plants and provides information on
24
The losing county entries in column (2) are weighted in the following manner. Losing counties are weighted by
the inverse of their number in that case. Losing plants are weighted by the inverse of their number per-county,
multiplied by the inverse of the number of losing counties in their case. The result is that each county (and each
plant within each county) is given equal weight within the case and then all cases are given equal weight.
147
some of their characteristics. In light of our sample selection criteria, the number of plants is of
special interest. On average, there are 18.8 plants in the winner counties and 25.6 in the loser
ones (and just 8.0 in the United States). The covariates are well balanced between plants in
winning and losing counties; in fact, there are no statistically significant differences either among
all plants or among the plants within the same 2-digit industry. 25
Overall, Table 3 has demonstrated that the MDP winner-loser research design balances
many (although not all) observable county-level and plant-level covariates. In the subsequent
analysis, we demonstrate that trends in TFP were similar in winning and losing counties prior to
the opening of the MDP, which lends further credibility to this design. Of course, this exercise
does not guarantee that unobserved variables are balanced across winner and loser counties or
their plants. The next section outlines our full econometric model and highlights the exact
assumptions necessary for consistent estimation.
IV. Econometric Model
Building on the model in section I, we start by assuming that incumbent plants use the
following Cobb-Douglas technology:
(6)
Ypijt = Apijt L1pljt K B 02pijt K
E
3pijt M
4pijt
where p references plant, i industry, j case, and t year; Yet is the total value of shipments; Apijt is
TFP; and we allow total labor hours of production (Lpijt) building capital stock (KBijt
machinery and equipment capital stock KEijt), and the dollar value of materials Mpijt
)
to have
separate impacts on output. In practice, the two capital stock variables are calculated with the
permanent inventory method that uses earlier years of the data on book values and subsequent
investment.26
Roughly 20% of the winners were in the Rust Belt, compared to roughly 25% of the losers (where the Rust Belt is
defined as MI, IN, OH, PA, NJ, IL, WI, NY). Roughly 65% of the winners were in the South, compared to roughly
45% of the losers.
26 For the first date available, plants' historical capital stock book values are deflated to constant dollars using BEA
data by 2-digit industry. In all periods, plants' investment is deflated to the same constant dollars using Federal
Reserve data by 3-digit industry. Changes in the capital stock are constructed by depreciating the initial deflated
capital stock using Federal Reserve depreciation rates and adding deflated investment. In each year, productive
capital stock is defined as the average over the beginning and ending values, plus the deflated level of capital rentals.
The analysis is performed separately for building capital and machinery capital. This procedure is described further
by Becker et al. (2005), Chiang (2004), and Davis et al. (1996), from whose files we gratefully obtained deflators.
25
148
Recall that equation (1) in Section I allows for agglomeration spillovers by assuming that
TFP is a function of the number of firms that are active in a county: Apjt = A(Npijt). Here we also
allow for some additional heterogeneity in Apijt. In particular, we generalize equation (1) by
allowing for permanent differences in TFP across plants (ap), cases (Xj), industry-specific timevarying shocks to TFP (tit), and a stochastic error term (Fput):
ln(Apljt) = ap + -t + Xj + Epljt + A(Nplt).
The goal is to estimate the causal effect of winning a plant on incumbent plants' TFP. To
do so, we need to impose some structure on A(Npijt). In particular, we use a specification that
allows for the new plant in winning counties to affect both the level of TFP as well as its growth
over time:
ln(Apit) = 6 1(Winner)j + yxtrendjt + Q (trend * 1(Winner))pjt
(7)
+ K 1(c >= O)jt + y (trend * l(z >= 0))jt
+ 01 (l(Winner) * l(T >= 0))pjt + 02 (trend * l(Winner) * l(u >= 0))pjt
+ ap+
Lt +
i + Epijt
where 1(Winner) is a dummy equal to 1 if plant p is located in a winner county; and t denotes
year, but it is normalized so that for each case the assigned year of the plant opening is C= 0.
The variable trendjt is a simple time trend.
Combining equations (6) and (7) and taking logs, we obtain the regression equation that
forms the basis of our empirical analysis:
(8)
lnYpt )=
n(Lpt )+ ln(KpBit
)+
n(pEt )+ 41n(Mpit
+ 8 1(Winner)p, + yl trendjt + 2 (trend * 1(Winner))pjt + K 1(t >= 0))jt + y (trend *
1( >=0O))jt
+ 01 (l(Winner) * 1(T >= 0))pjt
+
ap +
+ 02
(trend * 1(Winner) * l(t >= 0))pjt
it + kj + Epijt.
Equation (8) is an augmented Cobb-Douglass production function that allows labor, building
capital, machinery capital, and materials to have differential impacts on output. The paper's
focus is the estimation of the impact of the new plant on incumbent plants' TFP, so the
parameters of interest are 01 and 02 , which are the spillover effects. The former tests for a mean
shift in TFP among incumbent plants in the winning county after the opening of the MDP, while
the latter allows for a trend break in TFP among the same plants.
149
In practice, we estimate two variants of Equation (8). In some specifications, we fit a
parsimonious model that simply tests for a mean shift. In this model, any productivity effect is
assumed to occur immediately and to remain constant over time. Specifically, we make the
restrictions that y =
= y = 02 = 0, which assumes that differential trends are not relevant here.
This specification is essentially a difference in differences estimator and we refer to it as Model
1. Formally, after adjustment for the inputs, 1(Winner),, and 1(z >= O)jt, the consistency of 01 in
this model requires the assumption that E[(1(Winner) * l(z >= 0))pjt Epjt up, tL,, Xj ] = 0.
In other specifications, we estimate the model without imposing such restrictions on the
trends. In other words, we estimate the entire Equation (8). We label this Model 2. While Model
1 only allows for a mean shift in productivity, this specification allows both for a mean shift and
a trend break in productivity. In other words, Model 2 allows us to investigate whether any
productivity effect occurs immediately and whether the impact evolves over time.
specification is demanding of the data, because our sample is only balanced through
t
This
= 5 so
there are only 6 years per case to estimate 01 and 02.27
Equation (8) allows for unobserved determinants of TFP that are unrelated to the plant
opening but could be confounded with the spillover effects if not properly accounted for. It
includes a differential intercept for all observations from winning counties, 6. It also allows for a
common time trend, yx. The parameter 2 allows the time trend to differ for winning counties
prior to the plant opening. This will serve as an important way to assess the validity of this
research design. Finally, K and y capture the level change and trend break in TFP common to
plants in winning and losing counties after the MDP opening (i.e., when z >= 0).
In addition, all our models include three sets of fixed effects. First, they include separate
fixed effects for each plant, ap, so the comparisons are within a plant. Second, ttt represents the
parameters associated with a vector of 2-digit SIC industry by year fixed effects to account for
industry-specific shocks to TFP. Third, the X,'s are separate fixed effects for each case that
ensure that the impact of the MDP's opening is identified from comparisons within a winnerloser pair; they are a way to retain the intuitive appeal of pairwise differencing in a regression
framework.
A few further estimation details bear noting. First, unobserved demand shocks are likely
27 This
specification allows for spillovers to affect the level of TFP and to grow over time as in Glaeser and Mare
(2001).
150
to affect input utilization, and this raises the possibility that the estimated p's are inconsistent
(see, e.g., Griliches and Mairesse 1995). This has been a topic of considerable research and we
are unaware of a bullet-proof solution. We implement the standard fixes including modeling the
inputs with alternative functional forms (e.g., the translog), using cost shares at the plant and
industry-level rather than estimating the P's, and controlling for flexible functions of investment,
capital and materials (Syverson 2004; van Biesebroeck 2004; Olley and Pakes 1996; Levinsohn
and Petrin 2003). Additionally, we experiment with adding fixed effects for region by year or
region by industry by year, and allowing the effect of inputs to differ by industry or by winnerand post-MDP status. The basic results are unchanged by these alterations in the specification.
We also note that unobserved demand shocks are only a concern for the consistent estimation of
the parameters of interest (i.e., 01 and 02) if they systematically affect incumbent plants in
winning counties in the years after the MDP's opening, after adjustment for the rich set of
covariates in equation (8).
Second, in some cases this equation is fit on a sample of plants from the entire country,
but in most specifications the sample is limited to plants from winning and losing counties in the
ASM for every year from u = -8 through r = 0. When data from the entire country is used, the
sample is limited to plants that are in the ASM for at least 14 consecutive years. The smaller
sample of plants from the winning and losing counties allows for the impact of the inputs and the
industry shocks to differ in the winning-losing county sample from the rest of the country.
Finally, for most of the analysis, we further restrict the sample to observations in the years
between - = -7 and c = 5. Due to the dates of the MDP openings, this is the longest period for
which we have data from all cases.28
Third, we probe the validity and robustness of our estimates with a number of alternative
specifications. For example, we investigate whether changes in the price of output, capital
utilization, public investment, or attrition influence the estimates.
Fourth, all of the reported standard errors are clustered at the county level to account for the
correlation in outcomes among plants in the same county.2
Fifth, we focus on weighted versions of equation (8). Specifically, the specifications are
28 Data
29
from all cases is also available for c= -8, but output in this period is used to weight the regressions.
We experimented with clustering the standard errors at the 2-digit SIC by county level, but this occasionally
produced variance-covariance matrices that weren't positive definite. In instances where they were positive definite,
these standard errors were similar to those from clustering at the county level.
151
weighted by the square root of the total value of shipments in -c = -8 to account for
heteroskedasticity associated with differences in plant size. This weighting also means that the
results measure the change in productivity for the average dollar of output, which in our view is
more meaningful than the impact of the MDP on the average plant.
The analysis will also explore two additional issues. It will report on the fitting of
versions of equation (8) that interact the spillover variables with measures of economic distance
between the MDP and the incumbent plant. These specifications assess whether the magnitude
of the estimated spillovers varies with economic distance. Finally, the paper will assess whether
the MDP's opening affects local skill adjusted wages and the entry and exit decisions of plants in
the MDP's county.
V. Results
This section is divided into four subsections. The first reports baseline estimates of the
effect of the opening of a new Million Dollar Plant on the productivity of incumbent plants in the
same county through the estimation of equation (8).
The second subsection discusses the
validity of our design and explores the robustness of our estimates to a variety of different
specifications. The third subsection explores potential channels for the agglomeration effects by
testing whether the estimated spillovers vary as a function of economic distance. The final
subsection discusses the implications of our estimates for the profits of local firms.
A.
Baseline Estimates
Columns (1) and (2) of Table 4 report estimated parameters and their standard errors
from a version of equation (8).
Specifically, the natural log of output is regressed on the natural
log of inputs, year by 2-digit SIC industry fixed effects, plant fixed effects, and the event time
indicators in a sample that is restricted to the years r = -7 through ' = 5. The parameters
associated with the event time indicators report mean TFP in winning and losing counties,
respectively, in each event year relative to the year before the MDP opened (i.e., we have
subtracted off the - = -1 parameter estimates for the winners (losers) from the winner (loser)
estimates for each event year). Column (3) reports the difference between the estimated TFP
levels within each year.
The top panel of Figure 1 separately plots the mean TFP levels for winner and loser
152
counties (taken from columns (1) and (2) of Table 4) against t. The bottom panel of Figure 1
plots the difference in the estimated winner and loser coefficients against t.
Thus, it is a
graphical version of column (3) of Table 4.
Two important findings are apparent in these figures. First, the trends in TFP among
incumbent plants were very similar in the winning and losing counties in the years before the
MDP opening. In fact, a statistical test fails to reject that the trends were equal. This finding
supports the validity of our identifying assumption that incumbent plants in losing counties
provide a valid counterfactual for incumbents in winning counties.
Second, there is a sharp upward break in the difference in TFP between the winning and
losing counties beginning with the year that the plant opened. The top graph reveals that this
relative improvement is due to the continued decline in TFP in losing counties and a flattening
out of the TFP trend in winning counties. The figures also serve to underscore the importance of
the availability of losing counties as a counterfactual. For example, a naYve comparison of mean
TFP in winning counties before and after the opening would lead to the conclusion that the
opening had a small negative impact on incumbents' TFP. Overall, these graphs reveal much of
the paper's primary finding. This relative increase in TFP among incumbent plants in winning
counties will be confirmed throughout the battery of tests in the remainder of the paper.
Before proceeding, it is worth noting that the TFP of incumbent plants was declining in
both sets of counties in advance of the MDP plant opening. This finding may appear surprising,
because productivity increases over time in the overall economy. However, this downward trend
in TFP among large manufacturing plants appears to be a general phenomenon and is not
specific to plants in winning and losing counties. For example, we estimated augmented CobbDouglas production functions where the constant dollar total value of shipments is the dependent
variable and the covariates are the capital stock, labor, materials (also in constant dollars), 2-digit
industry fixed effects, and plant fixed effects. The equations are weighted by the total value of
shipments and the sample is restricted to plants in the ASM for at least 14 consecutive years in
the period 1973-1998. Thus, the approach and sample selection rule are similar in spirit to the
ones used in the MDP analyses.
The average 6 year change in TFP calculated over the years preceding the MDP openings
153
was a statistically significant decline of 4.7%. 30 This decline is similar to the decline among
incumbents in winning and losing counties in Figure 1. To the best of our knowledge, this
finding of declining TFP among large manufacturing plants has not been noted previously. 3 1
Now turning to the statistical models, the first four columns of Table 5 present the results
from fitting different versions of equation (8).
Models 1 and 2 are in Panels A and B,
respectively. Panel A reports the estimated mean shift parameter, 01, and its standard error (in
parentheses) in the "Mean Shift" row. Panel B reports the estimated impact of the MDP on
incumbent plants' TFP evaluated at c = 5 in the "Effect after 5 years" row, which is determined
by 01 and 02 that are also both reported. 32 The row "Pre-Trend" contains the coefficient
measuring the difference in the pre-existing trends between plants in the winning and losing
counties. In all of these specifications, the estimated impact of the MDP's opening is determined
during the period where -7 < z < 5, as the sample is balanced during these years.
In columns (1) and (2), the sample includes all manufacturing plants in the ASM that
report data for at least 14 consecutive years, excluding all plants owned by the MDP firm. In
column (3), the sample is restricted to include only plants in counties that won or lost a MDP.
This restriction means that the impact of the inputs and the industry-year fixed effects are
estimated solely from plants in these counties. Incumbent plants are now required to be in the
data only for -8 < z < 0 (not also for 14 consecutive years, though this does not change the
results).
Finally in column (4), the sample is restricted further to include only plant-year
observations within the period of interest (where r ranges from -7 through 5). This forces the
input parameters and industry-year fixed effects to be estimated solely on plant by year
observations that identify the spillover parameters.
This sample is used throughout the
remainder of the paper. Estimation details are noted at the bottom of the table and apply to both
Models 1 and 2.
The entries in Table 5 confirm the visual impression from Figure 1 that the opening of the
30
The 6 year average is a weighted averaged calculated over the 6 year periods before each of the MDP openings
where the weights are the number of plant openings associated with each 6 year period. For example, if there are 2
plant openings in 1987 and I in 1988, then the average change between 1980 and 1986 receives twice the weight as
the average change calculated between 1981 and 1987.
31 Since we are looking at large plants that have been active for a large number of years, we speculate that this
decline may have to do with aging. Additionally, many of the years preceding the MDP openings were in the late
1970s and early 1980s, which was a period of poor economic performance. Foster, Haltiwanger, and Krizan (2000)
have documented that within plant productivity growth is positively correlated with the economic cycle.
32 This is calculated as 01 + 602, because we allow the MDP to affect
outcomes fromr = 0 through r = 5.
154
MDP is associated with a substantial increase in TFP among incumbent plants in winning
counties. Specifically, Model 1 implies an increase in TFP of roughly 4.8%. As the figure
highlighted, however, the impact on TFP appeared to be increasing over time so Model 2 seems
more appropriate. This model's results suggest that the MDP's opening is associated with an
approximately 12% increase in TFP five years later. The estimates from both models would be
judged to be statistically different from zero by conventional criteria and are unaffected by the
changes in the specifications. Furthermore, the entries in the "Pre-trend" row demonstrate that
the null hypothesis of equal trends in TFP among incumbents in winning and losing counties
cannot be rejected.
The numbers in square brackets in column 4 measure the average size of the spillover
from a MDP opening in millions of 2006$.
This figure is calculated by multiplying the
estimated impact by the mean value of incumbent plants' total shipments in winning counties in
, = -1. This calculation indicates that the increase in TFP from a MDP was associated with an
increase in total output of about $170 million per year in Model 1. The Model 2 estimate is even
larger, suggesting an increase in output of roughly $429 million in year z = 5. These numbers
are large, with the Model 2 effect at - = 5 nearly the average level of MDP output.
Column (5) presents the results from a "naYve" estimator that is based on using plant
openings without an explicit counterfactual.
Specifically, a set of 47 plant openings were
randomly chosen from the ASM in the same years and industries as the MDP openings. The
remainder of the sample includes all manufacturing plants in the ASM, not also owned by the
randomly chosen firm, that report data for at least 14 consecutive years. With these data, we fit a
regression of the natural log of output on the natural log of inputs, year by 2-digit SIC fixed
effects, and plant fixed effects. In Model 1, two additional dummy variables are included for
whether the plant is in a winning county 7 to 1 years before the randomly chosen opening or 0 to
5 years after. The reported mean shift is the difference in these two coefficients (i.e., the average
change in TFP following the opening). In Model 2, the same two dummy variables are included
along with pre- and post-trend variables. The shift in level and trend are reported, along with the
pre-trend and the total effect evaluated after 5 years.
This naive "first-difference" style estimator indicates that the opening of a new plant is
associated with a -6% to -8% effect on incumbent plants' TFP, depending on the model. If the
estimates from the MDP research design are correct, then this naYve approach understates the
155
extent of spillovers by 13% (Model 1) to 18% (Model 2). Interestingly, the parameter on the
"pre-trend" indicates that the TFP of the incumbent plants was on a downward trend in advance
of the openings in the counties that attracted these new plants.
This is similar to what is
observed in our MDP sample of winners. Overall, the primary message is that the absence of a
credible research design can lead to misleading inferences in this setting.
It is natural to wonder about the degree of heterogeneity in the treatment effects from the
47 separate case studies that underlie the estimates presented thus far. Figure 2 explores this
heterogeneity by plotting case-specific estimates of parameter 01 in Model 1 and their 95%
confidence interval. Specifically, the Figure plots results from a version of Model 1 that interacts
the variable (l(Winner) * l(z >= 0)) with indicators for each of the cases. There are 45 estimates
of 01, one for each case. (Results from two cases were omitted to comply with the Census
Bureau's confidentiality rules). The figure reveals that there is heterogeneity in the estimated
impacts on TFP of incumbent plants. 27 of the 45 estimates are positive. 13 of the positive
estimates would be judged to be statistically different from zero at the 5% level, while the
comparable figure for the negative estimates is 9.
We assessed whether the estimated spillover effects are related to characteristics of the
MDP. Specifically, we regressed the estimates against three measures of the MDP's size, whether
the MDP is owned by a foreign company, and whether it is an auto company. When these
multiple measures were included jointly, none were significantly related to the estimated effect
of the MDP's opening. 33
Ultimately, TFP is a residual and residual labeling must be done cautiously.
As an
alternative way to examine the impact of the MDP on incumbents' productivity, we have
estimated directly the impacts of a MDP opening on output (unadjusted for inputs) and inputs.
The intent is to contrast the changes in outputs and inputs to shed light on whether productivity
increased without imposing the structural assumptions of the production function. Put another
way, are the incumbents producing more with less after the MDP's opening? Appendix Table 1
reports on estimates of the impact of a MDP opening on incumbents' output and usage of inputs.
33 Separate regressions of the case specific effects on the MDP's total output or the MDP's total labor force generated
statistically significant negative coefficients. This result is consistent with the possibility that when the MDP is very
large incumbents are left to hire labor and other inputs that are inferior in unobserved ways. On the other hand, we
failed to find any significant differences when separately testing whether the productivity effect varied by the ratio
of the MDP's output to county-wide manufacturing output, whether the MDP is owned by a foreign company, or
whether the MDP is an auto company.
156
The estimates are from the Model 1 and Model 2 versions of equation (8) with the key difference
that these equations do not include the inputs as covariates. Again, we use the Model 2 results to
estimate the impact of the opening 5 years afterwards.
Column (1) reports that the MDP opening is associated with an 8-12% increase in output.
Columns (2) through (5) report the results for the four inputs. It is striking that the change in all
of the inputs is roughly equal to or less than the increase in output. The Model 2 results are
especially noteworthy, because the 8% increase in output is accompanied by no increase in either
form of capital. Overall, the results in Appendix Table 1 indicate that, after the MDP's opening,
incumbent plants produced more with less; that is, they suggest that these plants became more
productive, and this is consistent with the TFP increases uncovered in Table 5.34
B.
Threats to Validity
Estimates in Table 5 appear to be consistent with significant agglomeration spillovers
generated by MDP openings. Although the comparisons in Table 3 and the similarity of the preexisting trends in TFP in winning and losing counties support the validity of the research design,
it is, of course, possible that there is a form of unobserved heterogeneity that accounts for the
higher levels of TFP in winning counties after the MDP's opening. Consequently, this subsection
investigates several possible alternative interpretations of the estimated spillover effects and
explores the robustness of the estimates to a variety of different assumptions. Specifically, we
investigate (i) the role of functional form assumptions and the possible presence of unobserved
industry and regional shocks; (ii) the endogeneity of inputs; (iii) changes in the price of output;
(iv) the possible role of public investment; (v) changes in capital utilization; and (vi) attrition.
(i) Functional Form, Industry Shocks and Regional Shocks. Table 6 reports on a
series of specification checks. For convenience, column (1) reports the results from the preferred
specification in column (4) specification of Table 5. These estimates are intended to serve as a
basis of comparison for the estimates in the remainder of the table.
We begin by generalizing our assumption on technology. Estimates in Table 5 assume a
Cobb-Douglas technology. In column (2) of Table 6, the inputs are modeled with the translog
functional form. Column (3) is based on a Cobb-Douglas technology but allows the effect of
34
Column (6) of Appendix Table 1 presents evidence on changes in the capital/labor ratio. The model suggests that
firms should substitute away from labor and toward capital. The estimated change in the capital/labor ratio is poorly
determined, making definitive conclusions unwarranted, but the point estimate is not supportive of this prediction.
157
each production input to differ at the 2-digit SIC level. This model accounts for possible
differences in technology across industries, as well as for possible differences in the quality of
inputs used by different industries.
For example, it is possible that even if technology was
similar across different manufacturers, some industries use more skilled labor than others.
Column (4) allows the effect of the inputs to differ in winning/losing counties and before/after
the MDP opening.
Columns (5) and (6) add census division by year fixed effects and census division by year
by 2-digit industry fixed effects.
These specifications aim to purge the spillover effects of
unobserved region-wide shocks or region by industry shocks to productivity that might be
correlated with the probability of winning a MDP.
Taken together, the results in columns (2) through (6) of Table 6 are striking. The
estimates appear to be insensitive to the specific functional form of the production function.
None of the specifications contradict the findings from the baseline specification in Table 5.
Although many of the estimates are smaller than the baseline ones, the magnitude of the decline
is modest. For example, they are all within one standard error of the baseline estimate in both
Models 1 and 2.
Overall, these results fail to undermine the conclusion from Table 5 that the
opening of a MDP leads to a substantial increase in TFP among incumbent plants and this is
consistent with theories of spillovers.
(ii) Endogeneity of Inputs. An important conceptual concern is that capital and labor
inputs should be treated as endogenous, because the same forces that determine output also
determine a firm's optimal choice of inputs (Griliches and Mairesse 1995). Unlike the usual
estimation of production functions, our aim is the consistent estimation of the spillover
parameters, 01 and 02, so the endogeneity of capital and labor is only relevant to the extent that it
results in biased estimates of these parameters.
This subsection employs the productivity
literature's state of the art techniques to control for the endogeneity of capital and labor to assess
this issue's relevance in this paper's setting.
We do this in two ways. First, in columns (7) and (8) of Table 6, we calculate TFP for
each plant by fixing the parameters on the inputs at the relevant input's share of total costs (van
Biesebroeck 2004; Syverson 2004; Foster, Haltiwanger, and Syverson 2007). This method may
mitigate any bias in the estimation of the parameters on the inputs associated with unobserved
demand shocks. In these two columns, the cost shares are calculated at the plant level and the 3-
158
digit SIC industry level over the full sample, respectively. The estimated spillover effects are
largely insensitive to this restriction.
Second, Table 7 presents estimates based on the widely-used methodologies proposed by
Olley and Pakes (1996) and Levinsohn and Petrin (2003). These methods are based on the result
that, under certain conditions, adjustment for investment or intermediate inputs (e.g., materials)
will remove the correlation between input levels and unobserved shocks to output. For example,
the column (2) specification adds 4 th degree polynomial functions of log building investment and
log machinery/equipment
specification adds
4 th
investment to the baseline specification. 35
The column (3)
degree polynomials in the two types of log capital stocks to the column (2)
equation. The column (4) specification is even richer as it adds all the "own" interactions
between polynomials in current investment and capital (i.e., the building investment polynomial
is interacted with the building capital stock polynomial, but it is not also interacted with the
machinery/equipment polynomials for stocks or investment) to column (3). Column (5) adds a
4th degree function of log materials to the baseline specification. In the column (6) specification,
a
4 th
degree polynomial in materials is fully interacted with
4 th
degree polynomials in building
capital and machinery capital. Column (7) includes fourth-degree polynomials in log materials
and log investment and log capital stock for both types of capital (not interacted).
Finally,
column (8) includes the controls from the columns (4) and (6) specifications.
The estimated spillover effects in Table 7 are generally consistent with the findings from
the baseline specification. This finding holds in both Models 1 and 2. Overall, this exercise fails
to suggest that the possible endogeneity of labor and capital is the source of the estimated
productivity spillovers.
(iii) Changes in the Price of Output. Another concern is that the theoretically correct
dependent variable is the quantity of output. However, due to the data limitations faced by
virtually all of the rest of the productivity literature, the dependent variable in our models is the
value of output or price multiplied by quantity. Consequently, it is possible that the estimated
spillover effect reflects higher prices, instead of higher productivity.
We do not expect this to be a major factor in our context. First, the sample is comprised
of manufacturing establishments that generally produce nationally traded goods. Therefore, it is
35
The exact measure used is log(l+investment), so zero values are not dropped. The results are very similar when
including polynomial functions of the level of investment and a dummy variable for values equal to zero.
159
likely that in many cases the price of output is set at the national level, and has little to do with
what happens in the county where the goods are produced. In the extreme case of a perfectly
competitive industry that produces a nationally traded good, there should be no effect on prices.
Second, we tested whether the size of the estimated productivity effect is larger in
industries that are more regional or more concentrated. The idea is that if price increases are
possible, then they should be larger in industries that are more local and/or in industries that are
less competitive. Consider for example, the case of an industry that sells mainly at the local level
(e.g., cement). The opening of a large new plant may ultimately increase incumbents' demand by
raising local income (even though the initial effect on demand should be negative). If the
industry is not very competitive, the increase in demand may ultimately lead to price increases
for the incumbents' output.
To implement the test, we estimated a Model 1 version of equation (8) that interacts
l(Winner)p,, 1(r >= 0))jt, and (1(Winner) * 1(t >= 0))pjt with incumbents' industry-specific
measure of average distance traveled by output between production and consumption. We also
conducted a similar exercise where these same variables are interacted with a measure of the
incumbent's industry concentration. 36 These specifications fail to produce evidence that the
estimated spillover effects are larger in more local industries or more concentrated industries; in
fact, there is some evidence for larger effects on incumbent plants that ship their products
further. Our conclusion is that price increases do not appear to be the source of the estimated
spillover effects.
(iv) Public Investment.
State and local governments frequently offer substantial
subsidies to new manufacturing plants to locate within their jurisdictions. These incentives can
include tax breaks, worker training funds, the construction of roads, and other infrastructure
investments. It is possible that some of the public investment in infrastructure benefits firms
other than the beneficiary of the incentive package. For example, the construction of a new road
intended for a MDP plant may also benefit the productivity of some of the incumbent firms. If
the productivity gains we have documented are due to public investment, then it is inappropriate
to interpret them as evidence of spillovers.
36
The information on distance is from Weiss (1972). Distance varies between 52 and 1337 miles, with a mean
of
498. Examples of regional industries are: hydraulic cement, iron and steel products, metal scrap and waste tailings,
ice cream and related frozen desserts, and prefabricated wooden buildings. The information on industry
concentration is from the Bureau of Census ("Concentration Ratios", 2002).
160
To investigate this possibility, we estimated the effect of MDP openings on government
total capital expenditures and government construction expenditures with data from the Annual
Survey of Governments. In models similar to equation (8), we find that the opening of a MDP
plant is associated with statistically insignificant increases in capital and construction
expenditures. In fact, in most specifications the estimated impact of a MDP opening is negative
and statistically insignificant.
Even in the specifications that produce positive insignificant
estimates, there is no plausible rate of return that could generate a meaningful portion of the
productivity gains in winning counties. Based on these measures of public investment, it seems
reasonable to conclude that public investment cannot explain the paper's results.
(v) Changes in Capital Utilization. Another potential threat to validity is that incumbent
plants may respond to the MDPs by increasing the intensity of their capital usage. This could
happen if depressed counties where the existing capital stock had been used below capacity win
the MDPs and increase production simply by operating their capital stock closer to capacity. As
an indirect test of this possibility, we estimated whether the MDP's opening affected the ratio of
the dollar value of energy usage (which is increasing in the use of the capital stock) to the capital
stock. Column 7 of Appendix Table 1 reports small and insignificant changes in this measure.
Thus, we conclude that greater capacity utilization is unlikely to be the source of the findings of
productivity spillovers.
(vi) Attrition of Sample Plants. Differential attrition in the sample of incumbent plants
in winning and losing counties could contribute to the measured differential in productivity
trends among survivors after the MDP's opening. This attrition could result from plants shutting
down operations or from plants continuing operations but dropping out of the group of plants
that are surveyed with certainty as part of the ASM.37
The available evidence suggests that differential attrition is unlikely to explain the finding
of spillovers in winning counties. First, in the baseline sample (i.e., the one used to produce the
results in column (4) of Table 5 and in Tables 6 and 7), 72% of the winning county plants
operating in the year of the MDP opening were still in the sample at its end (i.e.,
t
= 5). The
analogous figure in losing counties is 68%. The slightly larger attrition rate in losing counties is
consistent with the paper's primary result. Specifically, one seemingly reasonable interpretation
Recall, establishments are sampled with certainty if they are part of a company with manufacturing shipments
exceeding $500 million or their total employment was at least 250.
37
161
of this result is that the MDP's opening allowed some winning county plants to remain open that
would have otherwise closed. Thus to the extent that a MDP opening keeps weaker plants
operating, the above analysis will underestimate the overall TFP increase.
Second, the
estimation of equation (8) on the sample of plants that is present for all years from -7 to +5 yields
results that are qualitatively similar to those from the full sample. Third, the null hypothesis of
equal trends in TFP among attriting plants in winning and losing counties prior to the MDP's
opening cannot be rejected; the TFP trend in winning counties minus the TFP trend in losing
counties was -0.0052 (0.0080). 38
C.
Estimates of Spillovers by Economic Distance
What can explain the productivity gains uncovered above? Section I A discussed some
possible mechanisms that may be responsible for agglomeration spillovers.
Tables 8 and 9
attempt to shed some light on the possible mechanisms by investigating how the measured
spillover effect varies as a function of economic distance.
By Industry. Table 8 shows separate estimates from the baseline model for samples of
incumbent plants in the MDP's 2-digit industry and all other industries. In general, one might
expect that the effects of spillovers decline with economic distance (equation 5'). Although
looking within-industry does not shed direct light on which channel is the source of the
spillovers, it seems reasonable to presume that spillovers would be larger within an industry.
In examining the 2-digit SIC MDP industry results, it is important to recall that just 16 of
the 47 cases have plants in the MDP's 2-digit industry.
We also note that there can be
substantial heterogeneity in technologies and labor forces among the industries within a 2-digit
SIC industry. However, this research design and the available data do not permit an examination
at finer industry definitions.
Column 1 of Table 8 repeats the all industries estimates from column (4) of Table 5 and
is intended to serve as a basis of comparison. Columns (2) and (3) report on estimates from the
In addition to the specification checks described in this section, we also tested whether the results are sensitive to
the choice of the date of the MDP's opening. The estimated spillovers are virtually unchanged when we use the year
that the plant is first observed in SSEL as the MDP's opening date. When the year of the MDP article in Site
Selection is used as the plant's opening date, the Model 1 results are nearly identical to those in the Table 5 column
(4) specification and roughly 5% in the Model 2 specification. When the estimating equation is unweighted, the
evidence in favor of spillover effects is weaker indicating that the spillovers are concentrated among the larger
plants in the sample. As discussed above, our view is that the economically relevant concept of spillover is the
change in productivity for the average dollar of output, rather than the average plant.
38
162
baseline specification for incumbent plants in the MDP's 2-digit industry and all other industries,
respectively. The entries in these columns are from the same regression. Just as in Table 5, the
numbers in square brackets convert the parameter estimates into millions of 2006$.
The impacts are substantially larger in the own 2-digit industry.
For example, the
estimated increase in TFP for plants in the same 2-digit industry is a statistically significant 17%
in Model 1 and a poorly determined 33% at - = 5 in Model 2. In contrast, the estimates for
plants in other industries are a statistically insignificant 3.3% in Model 1 and marginally
significant 8.9% in Model 2. These basic findings in the own 2-digit and other industries are
robust to the different specifications in tables 6 and 7.39
Figures 3 and 4 provide 2-digit MDP industry and other industry analogues to Figure 1.
Importantly, there is not evidence of differential trends in the years before the MDP's opening
and statistical tests confirm this visual impression. The 2-digit MDP industry estimates are noisy
due to the small sample size, which was also evident in the statistical results. Just as in Figure 1,
the estimated impact reflects the continuation of a downward trend in TFP in losing counties and
a cessation of the downward trend in winning counties.
By Direct Measure of Economic Proximity. Having found that the spillover is larger
for incumbent plants in the MDP industry, we investigate the role of economic proximity more
directly by using several explicit measures of economic proximity that capture worker flows,
technological proximity, and input-output flows. To ease the interpretation, the economic
proximity or linkage variables are standardized to have a mean of zero and standard deviation of
one. In all cases, a positive value indicates a "closer" relationship between the industries.
Specifically, we estimate the following equation:
(8')
ln(Ypit)
1ln(Ljt)b+
B+
2 ln(KBt
6 3 1n(KE )+f 41n(Mpit
+ 6 1(Winner)pj + K 1(u >= 0))jt + 01 (1(Winner) * 1(c >= 0))pjt
+ 7nl 1(Winner)j * Proximity, + 12 (I(T >= O)jt * Proximity,)
+ r 3 (1(Winner) * 1(z >= O)pt * Proximityij ) + ap +
tlt +
j + plJt
where Proximityij is a measure of economic proximity between the incumbent plant industry and
the MDP industry. This equation is simply an augmented version of Model 1 that adds
39
Within the same 2-digit SIC, 71% of incumbents in winning counties and 69% of incumbents in losing counties
were still in the sample 5 years after the opening. Additionally, attriting plants within the same 2-digit SIC were
also on statistically indistinguishable trends prior to the MDP opening. Thus, differential attrition seems unlikely to
explain the 2-digit results.
163
interactions of the industry linkage variables with 1(Winner)j, 1(t >= 0))jt , and (1(Winner) * 1(c
>= 0))pj,.
The coefficient of interest is n 3 , which is the coefficient on the triple interaction
between the dummy for winner, the dummy for "after," and the measure of proximity. This
coefficient assesses whether "closer" industries benefit more from the MDP's opening.
A
positive coefficient means that the estimated productivity spillover is larger after the MDP
opening for incumbents that are geographically and economically close to the new plant, relative
to incumbents that are geographically close but economically distant from the new plant (relative
to the same comparison among incumbents in loser counties). A zero coefficient means that the
estimated productivity spillover is the same for all the incumbents in a county, regardless of their
economic proximity to the new plant.
Table 9 reports estimates ofn 3. The first 6 columns include the interactions in one at a
time. For example, column (1) suggests that a one standard deviation increase in the CPS
Worker Transitions variable between incumbent plants' industry and the MDP's industry is
associated with a 7% increase in the spillover. This finding is consistent with the theory that
spillovers occur through the flow of workers across firms. One possibility is that new workers
share ideas on how to organize production or information on new technologies that they learned
with their previous employer. This measure tends to be especially high within 2-digit industries,
so this finding was foreshadowed by the own 2-digit results in Table 8.
The measures of intellectual or technological linkages indicate meaningful increases in
the spillover. The precise mechanism by which these ideas are shared is unclear, although both
the flow of workers across firms and the mythical exchange of ideas over beers between workers
from different firms are possibilities. Notably, there is more variation in these measures within 2digit industries than in the CPS labor transitions measure.
Columns (5) and (6) provide little support for the flow of goods and services in
determining the magnitude of spillovers. Thus, the data fail to support the types of stories where
an auto manufacturer encourages (or even forces) its suppliers to adopt more efficient production
techniques. Recall, all plants owned by the MDP's firm are dropped from the analysis, so this
finding does not rule out this channel within firms. The finding on the importance of labor flows
is consistent with the results in Ellison, Glaeser and Kerr (2007) and Dumais, Ellison, and
Glaeser (2002), while the finding on input and output flows stands in contrast with these papers'
findings.
164
In the column (7) specification, we include all the interactions simultaneously. The labor
flow, the citation pattern, and the technology input interactions all remain positive but now
would be judged to be statistically insignificant. The interactions with proximity to customers
and suppliers are now both negative.
Overall, this analysis provides some support for the notion that spillovers occur between
firms that share workers and between firms that use similar technologies. In terms of Section IC,
this evidence is consistent with intellectual externalities, to the extent that they are embodied in
workers who move from firm to firm, and to the extent that they occur among firms that use
technologies that are reasonably similar. Table 9 seems less consistent with the hypothesis that
agglomeration occurs because of proximity to customers and suppliers. We caution against
definitive conclusions, because the utilized measures are all imperfect proxies for the potential
channels. Further, the possibility of better matches between workers and firms could not be
directly tested with these data.
D.
Entry and Labor Costs as Indirect Tests of Spillovers
The paper has uncovered economically
sizable productivity gains for incumbent
establishments following the opening of the new MDP. For example, a MDP plant opening is
associated with a 12% increase in TFP five years later. This effect is even larger for incumbent
plants that are in the same 2-digit industry of the new plant and for plants that tend to share
workers and technologies. In the presence of positive spillovers, the model has two empirical
predictions which this subsection tests.
First, if the spillovers are of a sufficient magnitude (i.e., they are larger than the increase
in costs in the short run), the MDP's county should experience entry by new firms (relative to the
losing counties). Table 10 tests this prediction. The entries in Panel 1 come from regressions
that use data from the Census of Manufactures, which is conducted every five years.
The
dependent variables are the log of the number of establishments (column 1) and the log of total
manufacturing output (column 2) in the county, respectively. In both columns, all plants owned
by the MDP's firm are excluded from the dependent variable.
The sample is comprised of
observations from winning and losing counties only. The covariates include a full set of county
fixed effects, year fixed effects, case fixed effects, and an indicator for whether the observation is
from after the MDP's opening. The parameter of interest is associated with the interaction of
165
indicators for an observation from a winning county and the post-opening indicator, so it is a
difference in differences estimator of the impact of the MDP's opening. 40
Column (1) reports that the number of manufacturing plants increased by roughly 12.5%
in winning counties after the MDP plant's opening.
A limitation of this measure is that it
assumes that all plants are of an equal size. The total value of output is economically more
meaningful, because it treats an increase in output at an existing plant and a new plant equally.
As column (2) highlights, the opening of a MDP plant is associated with a roughly 14.5%
increase in total output in the manufacturing sector although this is not estimated precisely.
Overall, these results are consistent with the TFP results of substantial spillovers in that it
appears that the MDP attracted new economic activity to the winning counties (relative to losing
ones) in the manufacturing sector. 4 1 Presumably, this new activity located in the winning
counties to gain access to the spillovers.
The second prediction is that if the spillovers are positive, the prices of local inputs will
increase as firms compete for these factors of production. The most important locally supplied
input for manufacturing plants is labor. Column (3) in Panel 2 of Table 10 reports the results
from regressions of the log wage using data from the 1970, 1980, 1990, and 2000 Censuses of
Population from the winning and losing counties. 42 These data are preferable to the measure of
labor costs reported in the Census of Manufacturers (i.e., the aggregate wage bill for production
and non-production workers), which does not provide information on the quality of the labor
force (e.g., education and experience). Specifically, we estimate models for In(wage) and control
for dummies for interactions of worker age and year, age-squared and year, education and year,
sex and race and Hispanic and citizen, and case fixed effects. We also include indicators for
whether the observation is from a winning county, occurs after the MDP's opening, and the
Because data is available every 5 years, depending on the Census year relative to the MDP opening, the sample
years are 1 - 5 years before the MDP opening and 4 - 8 years after the MDP opening. Thus, each MDP opening is
40
associated with one earlier date and one later date. Models are weighted by the number of plants in the county in
years -6 to -10 and column 4 is weighted by the county's total manufacturing output in years -6 to -10.
41 It is possible that the MDP's spillovers extended beyond manufacturing. In this case, it might be reasonable to
expect increased entry in other sectors too.
42 The sample is limited to individuals who worked last year, worked more than 26 weeks, usually work more than
20 hours per week, are not in school, are at work, and who work for wages in the private sector. One important
limitation of the Census data is that they lack exact county identifiers for counties with populations below 100,000.
Instead, it is possible to identify PUMAs in the Census, which in rural areas can include several counties. This
introduces significant measurement error, which is partly responsible for the imprecision of the estimate.
166
interaction of these two indicators. 43 This interaction is the focus of the regression and is an
adjusted difference in differences estimator of the impact of the MDP's opening on wages. This
equation is analogous to the Model 1 version of equation (8) that was used to analyze TFP.
The estimate indicates that after adjusting for observable heterogeneity, wages increase
by 2.7% in winning counties after the MDP's opening. This effect appears quantitatively sizable
and is marginally statistically significant.
The multiplication of the estimated 2.7% wage
increase by the average labor earnings in winning counties implies that the quality-adjusted
annual wage bill for employers in all industries increased by roughly $151 million after the
MDP's opening. This finding is consistent with positive spillovers and an upward sloping labor
supply curve, perhaps due to imperfect mobility of labor (as in Section I).
It is possible to use the estimated increase in wages to make some back of the envelope
calculations of the MDP's impact on incumbent plants' profits. Recall, the Model 1 result in
Table 5 indicated an increase in TFP of approximately 4.8% (we focus on Model 1 because it is
impossible to estimate a version of Model 2 with the decennial population Census data). If we
assume that workers are homogenous or that high and low skill workers are perfectly
substitutable in production, then the labor market-wide increase in wages applies throughout the
manufacturing sector. In our sample, labor accounts for roughly 23% of total costs, so the
estimated 2.7% increase in skill adjusted wages implies that manufacturers' costs increased by
approximately 0.62%. The increased production costs due to higher wages are therefore 13% of
the gain in TFP.
These calculations demonstrate that the gains in TFP do not translate directly to profits
due to the higher costs of local inputs. Since the prices and quality of other inputs are not
observable, it is not possible to determine the total increase in production costs. Further, the
observations on wages occurs only a few years after the MDP's opening and the impact on
wages may increase more as production expands to gain access to the spillover. For these
reasons, this back of the envelope calculation should be interpreted as a lower bound of the
increase in input costs. In the long run, an equilibrium requires that the total impact on profits is
zero.
The pre-period is defined as the most recent census before the MDP opening. The post-period is defined as the
most recent census 3 or more years after the MDP opening. Thus, the sample years are 1 - 10 years before the MDP
opening and 3 - 12 years after the MDP opening.
43
167
VI. Summary and Interpretation
This paper makes three principal contributions. This section summarizes them and places
them in some context.
The first and most robust result is that the successful attraction of a "Million Dollar
Plant" is on average associated with increased TFP for incumbent plants. This finding is robust
to a battery of specification tests.
In the preferred Model 2, the estimates suggest that
incumbents' TFP in winning counties was about 12% higher. This translates into an additional
$430 million in annual output five years after the MDP's opening. This is an economically large
number and it is natural to wonder whether it is "too large" to be plausible.
There are several related issues worth noting when considering this possibility. First,
there may be an unobserved factor correlated with the MDP's opening that can explain these
results and would invalidate the identifying assumption. The likelihood of this possibility is
diminished by the similarity of the pre-trends in TFP documented in Figure 1 and the balancing
of many of the ex-ante observable characteristics of winning and losing counties and their
incumbent plants. Nevertheless, this possibility cannot be dismissed as would be the case in a
randomized experiment.
Second, it is possible that the estimated impacts on TFP are influenced by changes in the
quality of the workforce employed by incumbent plants. The sign of this change is a priori
unclear. On the one hand, the MDP may attract higher quality workers to the county, allowing
the incumbents to upgrade the quality of their workforce. On the other hand, the MDP could hire
away the incumbents' best workers. Regardless of which force prevails, this could affect the
estimated impact on TFP because labor is measured as total hours by production workers (recall
that education or other measures of skill are not included in the ASM).
Third, the external validity of the results is unknown. In particular, the 47 MDP plants in
the sample differ from the average manufacturing plant in many respects.
Perhaps most
importantly, these plants generated bidding from local governments, which is a first indication
that there may be an ex-ante expectation of significant spillovers.
Further, these plants are
substantially larger than the typical manufacturing plant. The point is that these results are
unlikely to generalize to the opening of more typical plants.
Fourth, the estimated impact on TFP is unlikely to be a structural estimate of the MDP's
impact on the TFP of incumbents. As Table 10 indicated, the MDP's opening appears to lead to
168
other plants opening in the same county. Thus, the estimated spillover may reflect spillovers
from the MDP and these new plants. Consequently, it is likely more appropriate to interpret the
TFP effect as the reduced form effect of the MDP's opening and everything that accompanies it
rather than the impact of the MDP alone.
The second contribution is a theoretical one that is supported by the data. In particular,
the model underscored that increases in TFP do not translate directly into higher profits available
for other manufacturing plants.
The model predicts that the increases in TFP will be
accompanied by increases in local input prices and that these increases are necessary for an
equilibrium. The finding of higher prices for quality-adjusted labor in Table 10 is consistent
with this prediction. Further, the increased levels of economic activity also documented in Table
10 reflect the increased demand to locate in the winning county that leads to higher local prices
and the new equilibrium.
The third contribution is that the paper has shed some light on the channels that underlie
the estimated spillovers. Specifically, our tentative conclusion is that the spillovers are larger
between firms that share workers and use similar technologies.
This is consistent with
intellectual externalities, to the extent that they are embodied in workers who move from firm to
firm and occur among firms that use technologies that are reasonably similar. Additionally, it is
consistent with higher rates of TFP due to improved efficiencies of worker-firm matches.
Finally, these results may have some surprising policy implications. A standard critique
of local governments providing subsidies to new plants to locate within their jurisdictions is that
it may be rational for the locality but it is welfare decreasing for the nation. Although the
economics of this argument are not always transparent, we believe that it generally refers to
cases of disequilibrium where local factors of production (e.g., labor or buildings) are
unemployed. In this case, tax competitions may be beneficial locally but suboptimal nationally
(at least in the absence of significant worker moving costs).
In contrast, the finding of spillovers is the basis of an economic justification for local
subsidies, even from a national perspective. Specifically, the MDP plant cannot capture these
spillovers on its own and consequently might choose a location where its costs are low but the
spillovers are minimal. However, the socially efficient outcome is for these plants to locate
where the sum of their profits and the spillovers are greatest. In this setting, national welfare is
maximized when payments are made to plants that produce the spillovers so that they internalize
169
this externality in making their location decision.
In thinking about the policy implications, it is important to bear in mind that the
estimated 12% gain in TFP is an average effect. As Figure 2 demonstrated, there is substantial
variability. For example, the estimated impact is negative in 40% of the cases. The point is that
risk averse local governments may be unwilling to provide tax incentives with this distribution of
outcomes.
VII.
Conclusion
Overall, this paper has documented that there are substantial spillovers flowing from
large new plants to incumbent plants. Further, these spillovers are larger between plants that
share labor pools and use similar technologies.
Thus, this paper has provided evidence
consistent with the idea that firms agglomerate in certain localities, at least in part, because they
are more productive for being close to other firms.
There are several implications for future research. First, the paper has demonstrated the
value of quasi-experiments that plausibly avoid the confounding of spillovers with differences in
the determinants of TFP across locations. Second, the paper highlights that tests for the presence
of spillovers can be conducted by directly measuring TFP. These tests can serve as an important
complement to measurement of rates of coagglomeration that may reflect spillovers, cost
shifters, or natural advantages. In this spirit, it is important to determine whether impacts on
TFP are evident outside the manufacturing sector. Third, the heterogeneity in the estimated
spillovers documented in Figure 2 and the results on the mechanism in Table 9 underscore that
there is still much to learn about the structural source of these spillovers. This is an especially
fruitful area for future research.
170
References
Acemoglu, Daron. 1996. "A Microfoundation for Social Increasing returns in Human Capital
Accumulation." QuarterlyJournalof Economics, 11 (August): 779-804.
Audretsch, David and Maryann Feldman. 1996. "R&D Spillovers and the Geography of Innovation and
Production." American Economic Review 86 (June): 630-640.
Audretsch, David and Maryann Feldman. 2004. "Knowledge Spillovers and the Geography of
Innovation." In Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and
Jacques-Frangois Thisse. Amsterdam: North-Holland.
Becker, Randy, John Haltiwanger, Ron Jarmin, Shawn Klimek, and Daniel Wilson. 2005. "Micro and
Macro Data Integration: The Case of Capital." Bureau of the Census: Centerfor Economic Studies
Working Paper05-02.
Bloom, Nick, Mark Schankerman, and John Van Reenen. 2007. "Identifying Technology Spillovers and
Product Market Rivalry." National Bureau of Economic Research Working Paper No. 13060.
Branstetter, Lee G. 2001. "Are Knowledge Spillovers International or Intranational in Scope?
Microeconometric
evidence from US and Japan." Journal of International Economics 53
(February): 53-79.
Card, David, Kevin F. Hallock, and Enrico Moretti. 2007. "The Geography of Giving: The Effect of
Corporate Headquarters on Local Charities." University of California, Berkeley Mimeograph.
Chiang, Hyowook. 2004. "Learning By Doing, Worker Turnover, and Productivity Dynamics."
Econometric Society 2004 Far Eastern Meeting Paper No. 593.
Davis, James and J. Vernon Henderson. 2004. "The Agglomeration of Headquarters." Bureau of the
Census: Centerfor Economic Studies Working Paper04-02.
Davis, Steven J., John C. Haltiwanger, and Scott Schuh. 1996. Job Creationand Destruction.Cambridge,
MA: MIT Press.
Davis, Donald and David E. Weinstein. 2002. "Bones, Bombs, and Break Points: The Geography of
Economic Activity." American Economic Review 92 (December).
Dumais, Guy, Ellison, Glenn, and Glaeser, Edward L. 2002. "Geographic Concentration as a Dynamic
Process." Review ofEconomics and Statistics 84 (May): 193-204.
Duranton, Giles and Diego Puga. 2004 "Micro-Foundations of Urban Agglomeration Economies." In
Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and Jacques-
Francois Thisse. Amsterdam: North-Holland.
Ellison, Glenn and Edward L. Glaeser. 1997. "Geographic Concentration in U.S. Manufacturing
Industries: A Dartboard Approach." Journalof PoliticalEconomy 105 (October): 889-927.
Ellison, Glenn and Edward L. Glaeser. 1999. "The Geographic Concentration of Industry: Does Natural
Advantage Explain Agglomeration?" American Economic Review 89 (May): 311-316.
Ellison, Glenn, Edward L. Glaeser, and William Kerr. 2007. "What Causes Industry Agglomeration?
Evidence from Coagglomeration Patterns." National Bureau of Economic Research Working Paper
No. 13068.
Foster, Lucia, John C. Haltiwanger, and Chad Syverson. 2007. "Reallocation, Firm
or
Profitability?"
Productivity
on
Selection
Efficiency:
and
Turnover,
University of Chicago Mimeograph..
Foster, Lucia, John C. Haltiwanger, and C.J. Krizan. 2000. "Aggregate Productivity Growth: Lessons
from Microeconomic Evidence." Mimeograph.
Glaeser, Edward L. 1999. "Learning in Cities." Journalof Urban Economics 46(September): 254-77.
Glaeser, Edward L. 2001. "The Economics of Location-Based Tax Incentives." HarvardInstitute of
Economic Research, Discussion Paper Number 1932.
Glaeser, Edward L. and Janet E. Kohlhase. 2003. "Cities, Regions and the Decline of Transport Costs."
Papers in Regional Science 83 (January): 197-228.
Glaeser, Edward L. and David C. Mare. 2001. "Cities and Skill." Journalof Labor Economics 19(2):
316-342.
171
Greenstone, Michael and Enrico Moretti. 2004. "Bidding for Industrial Plants: Does Winning a 'Million
Dollar Plant' Increase Welfare?" National Bureau of Economic Research Working Paper No. 9844.
Griliches, Zvi. 1958. "Research Cost and Social Returns: Hybrid Corn and Related Innovations." Journal
of PoliticalEconomy 66 (October): 419-431.
Griliches, Zvi and Jacques Mairesse. 1995. "Production Functions: the Search for Identification."
National Bureau of Economic Research Working Paper No. 5067.
Grossman, Gene M. and Elhanan Helpman. 1991. Innovation and Growth in the Global Economy.
Cambridge, MA: MIT Press.
Henderson, J. Vernon. 2001. "Urban Scale Economies." In Handbook of Urban Studies, edited by Ronan
Paddison. London: SAGE Publications.
Henderson, J. Vernon. 2003. "Marshall's Scale Economies." Journal of Urban Economics 55 (January):
1-28.
Henderson, J. Vernon and Duncan Black. 1999. "A Theory of Urban Growth." Journal of Political
Economy, 107 (April): 252-284
Jaffe, Adam B. 1986. "Technological Opportunities and Spillovers of R&D: Evidence from Firm's
Patents, Profit and Market Value." American Economic Review 76 (December): 984-1001.
Jaffe, Adam B. and Manuel Trajtenberg. 2002. Patents, Citations and Innovations: A Window on the
Knowledge Economy. Cambridge, MA: MIT Press.
Jaffe, Adam B., Manuel Trajtenberg, and Michael S. Fogarty. 2000. "Knowledge Spillovers and Patent
Citations: Evidence from a Survey of Inventors." American Economic Review 90 (May): 215-218.
Jaffe, Adam B., Manuel Trajtenberg and Rebecca M. Henderson. 1993. "Geographic Localization of
Knowledge Spillovers as Evidenced by Patent Citation." Quarterly Journal of Economics 108
(August): 577- 598.
Jovanovic, Boyan and Rafael Rob. 1989. "The Growth and Diffusion of Knowledge." Review of
Economic Studies 56 (October): 569-582.
Krugman, Paul. 1991 a. Geography and Trade. Cambridge, MA: MIT Press.
Krugman, Paul. 1991b. "Increasing Returns and Economic Geography," Journalof PoliticalEconomy 99
(June): 483-499.
Levinsohn, James and Amil Petrin. 2003. "Estimating Production Functions Using Inputs to Control for
Unobservables." Review of Economic Studies 70 (April): 317-342.
Lucas, Robert E. Jr. 1988. "On the Mechanics of Economic Development." Journal of Monetary
Economics, 22 (July): 3-42.
Marshall, Alfred. 1890. Principles of Economics. New York: Macmillan and Co.
McGahan, Anita W. and Brian S. Silverman. 2001. "How Does Innovative Activity Change as
Industries Mature?" InternationalJournalofIndustrialOrganization 19 (July): 1141-1160.
Moretti, Enrico. 2004a. "Estimating the external return to higher education: Evidence from crosssectional and longitudinal data." JournalofEconometrics 120 (July-August): 175-212.
Moretti, Enrico. 2004b. "Workers' Education, Spillovers and Productivity: Evidence from PlantLevel Production Functions." American Economic Review 94 (June): 656-690.
Moretti, Enrico. 2004c. "Human Capital Externalities in Cities." In Handbook of Urban and
RegionalEconomics 4, edited by J. Vernon Henderson and Jacques-Frangois Thisse. Amsterdam:
North-Holland.
Mowery, David C. and Arvids A. Ziedonis. 2001. "The Geographic Reach of Market and Non-market
Channels of Technology Transfer: Comparing Citations and Licenses of University Patents."
National Bureau of Economic Research Working Paper No. 8568.
Olley, Steven G. and Ariel Pakes. 1996. "The Dynamics of Productivity in the
Telecommunications Equipment Industry." Econometrica64 (November): 1263-98.
Ottaviano, Gianmarco and Jacques-Frangois Thisse. 2004. "Agglomeration and Economic Geography." In
Handbook of Urban and Regional Economics 4, edited by J. Vernon Henderson and JacquesFrangois Thisse. Amsterdam: North-Holland.
Petrongolo, Barbara and Christopher A. Pissarides. 2005. "Scale Effects in Markets with Search." The
172
Economic Journal, 116 (January): 21-44.
Rauch, James E. 1993. "Productivity Gains from Geographic Concentration of Human Capital: Evidence
from the Cities." Journalof Urban Economics 34 (November): 380-400.
Roback, Jennifer. 1982. "Wages, Rents and the Quality of Life." Journalof PoliticalEconomy 90
(December): 1257-1278.
Roback, Jennifer. 1988. "Wages, rents and amenities: Differences among workers and regions."
Economic Inquiry 26 (January): 23-41.
Rosenbaum, Paul R. and Donald B. Rubin. 1983. "The Central Role of the Propensity Score in
Observational Studies for Causal Effects." Biometrika 70 (April): 41-55.
Rosenthal, Stuart S. and William C. Strange. 2001. "The Determinants of Agglomeration." Journalof
Urban Economics 50 (September): 191-229.
Rosenthal, Stuart S. and William C. Strange. 2004. "Evidence on the Nature and Sources of
Agglomeration." In Handbook of Urban and RegionalEconomics 4, edited by J. Vernon
Henderson and Jacques-Frangois Thisse. Amsterdam: North-Holland.
Saxenian, Anna Lee. 1994. RegionalAdvantage.: Culture and Competition in Silicon Valley and Route
128. Cambridge, MA: Harvard University Press.
Syverson, Chad. 2004a. "Market Structure and Productivity: A Concrete Example." Journalof Political
Economy 6 (December): 1181-1222.
Syverson, Chad. 2004b. "Product Substitutability and Productivity Dispersion." Review of Economics and
Statistics2 (May): 534-550.
van Biesebroeck, Johannes. 2004. "Robustness of Productivity Estimates." National Bureau of Economic
Research Working Paper No. 10303.
Venable, Tim. 1992. "BMW Drives into South Carolina with $300 Million Auto Plant." Site Selection
August: 630-631.
Weiss, Leonard W. 1972. "The Geographic Size of Markets in Manufacturing." The Review of Economics
and Statistics 54 (August): 245-257.
Zucker, Lynne G., Michael R. Darby, and Marilynn B. Brewer. 1998. "Intellectual Human Capital and the
Birth of U.S. Biotechnology Enterprises." American Economic Review 88 (March): 290-306.
173
Figure 1: The Effect of a "Million Dollar Plant" Opening on TFP of All Manufacturing Plants in Winner
and Loser Counties.
All Industries: Winners Vs. Losers
0.1
005
-7
-0.05
-7
-2
-6
-5
-4
-3
-6
-5
-4
-3
-1
-2
-1
0
-
3
.
4
-
-0.1
-0.15
Year, relative to opening
-
Winning Counties
--- A--- Losing Counties 1
1
Difference: Winners - Losers
0.1
0.05 -
/j
-0.05
-7
-6
-5
-3
-2
-1
0
-0.05
Year, relative to opening
Notes: These figures accompany Table 4.
174
1
2
3
4
5
Figure 2. Distribution of Case-Specific Mean Shift Effects from the Opening of a "Million Dollar Plant"
0.5
0.4
0.3
0.2
t
11
4
t
-0.2
S
,
-0.3
-0.4
-0.5
Notes: The figure reports results from a version of Model 1 that estimates the parameter 01 for each of the 47 MDP
cases. The figure reports only 45 estimates because two cases were dropped for Census confidentiality reasons.
175
Figure 3: The Effect of a "Million Dollar Plant" Opening on TFP of Manufacturing Plants in the MDP's
2-Digit Industry in Winner and Loser Counties.
2-digit MDP Industry: Winners Vs. Losers
0.1
A'
01
-
0A
-
-1
A.
9Q--'
3
-A...--------
4
........
A
Year, relative to opening
-----
Winning Counties
--- A--- Losing Counties
Difference (Winners - Losers)
0.3 0.2
-
0.1
0
-0.1
-7
-1
-6
0
-0.2
-0.3
Year, relative to opening
Notes: These figures accompany Table 8, Column 2 (MDP's 2-digit Industry).
176
1
2
3
4
5
Figure 4: The Effect of a "Million Dollar Plant" Opening on TFP of Manufacturing Plants in All
Industries, Except the MDP's 2-Digit Industry in Winner and Loser Counties.
Other Industries: Winners Vs. Losers
0.1
0.05
-7
-0.05
-6
-5
-4
-3
-2
-1
..
0
-.
-0.1
-0.15
Year, relative to opening
---
Winning Counties
Losing Counties
- - -A---
Difference (Winners - Losers)
0.1 -
0.05-
0
I
-7
-6
-5
"
-3
-2
Y
-
-1
0
1
2
3
4
-0.05
Year, relative to opening
Notes: These figures accompany Table 8, Column 3 (All 2-digit Industries, except the MDP's 2-digit Industry).
177
5
Table 1. The "Million Dollar Plant" Sample
(1)
Sample MDP Openings':
Across All Industries
Within Same 2-digit SIC
47
16
Across All Industries:
Number of Loser Counties per Winner County:
1
2+
31
16
Reported Year - Matched Year:
-2 to -1
0
1 to 3
2
Reported Year of MDP Location:
1981 -1985
1986 - 1989
1990- 1993
20
15
12
11
18
18
MDP Characteristics, 5 years after opening:3
Output ($1000)
452801
(901690)
0.086
Output, relative to county output 1 year prior
(0.109)
2986
Hours of Labor (1000)
(6789)
Million Dollar Plant openings that were matched to the Census data and for which there were incumbent plants in
both winning and losing counties that are observed in each of the eight years prior to the opening date (the opening
date is defined as the earliest of the magazine reported year and the year observed in the SSEL.) This sample is then
restricted to include matches for which there were incumbent plants in the Million Dollar Plant's 2-digit SIC in both
locations.
2 Only a few of these differences are 3. Census confidentiality rules prevent being more specific.
3 Of the original 47 cases, these statistics represent 28 cases. A few very large outlier plants were dropped so that
the mean would be more representative of the entire distribution (those dropped had output greater than half of their
county's previous output and sometimes much more). Of the remaining cases: most SSEL matches were found in
the ASM or CM but not exactly 5 years after the opening date; a couple of SSEL matches in the 2xxx-3xxx SICs
were never found in the ASM or CM; and a couple of SSEL matches not found were in the 4xxx SICs. The MDP
characteristics are similar for cases identifying the effect within same 2-digit SIC. Standard deviations are reported
in parentheses. All monetary amounts are in 2006 US dollars.
178
Table 2. Summary Statistics for Measures of Industry Linkages
Mean
Only 4th
Standard
Only 1st
Measure of
Deviation
Quartile
All Plants
Quartile
Description
Industry Linkage
Labor Market Pooling:
0.317
0.249
0.002
0.119
Proportion of workers leaving a job
CPS Worker
Transitions
in this industry that move to the
MDP industry (15 months later)
Intellectual or Technology Spillovers:
0.033
0.001
0.057
0.022
Percentage of manufactured industry
Citation pattern
patents that cite patents manufactured
in MDP industry
0.084
0.106
0.022
0.000
R&D flows from MDP industry, as a
Technology
Input
percentage of all private sector
technological expenditures
0.035
0.042
0.011
0.000
R&D flows to MDP industry, as a
Technology
Output
percentage of all original research
expenditures
Proximity to Customers and Suppliers:
0.061
0.075
0.017
0.000
Industry inputs from MDP industry,
Manufacturing
as a percentage of its manufacturing
Input
inputs
0.139
0.163
0.042
0.000
Industry output used by MDP
Manufacturing
industry, as a percentage of its output
Output
to manufacturers
Notes: CPS Worker Transitions was calculated from the frequency of worker industry movements in the rotating
CPS survey groups. This variation is by Census Industry codes, matched to 2-digit SIC. The last 5 measures of
cross-industry relationships were provided by Ellison, Glaeser, and Kerr (NBER Working Paper 13068). These
measures are defined in a 3-digit SIC by 3-digit SIC matrix, though much of the variation is at the 2-digit level. In
all cases, more positive values indicate a closer relationship between industries. Column 1 reports the mean value of
the measure for all incumbent plants matched to their respective MDP. Column 2 reports the mean for the lowest
25% and column 3 reports the mean for the highest 25%. Column 4 reports the standard deviation across all
observations. The sample of plants is all incumbent plants, as described for Table 1, for which each industry linkage
measure is available for the incumbent plant and its associated MDP. These statistics are calculated when weighting
by the incumbent plant's total value of shipments eight years prior to the MDP opening.
179
Table 3. County & Plant Characteristics by Winner Status, One Year Prior to a "Million Dollar Plant" Opening
All Plants
All US
Counties
(3)
(1) -(2)
t-stat
(4)
(1)- (3)
t-stat
(5)
Winning
Counties
(6)
16
20,628
11,259
-2.05
5.79
20,230
20,528
11,378
-0.11
4.62
0.074
0.096
0.037
-0.81
1.67
0.076
0.089
0.057
-0.28
0.57
322,745
447,876
82,381
-1.61
4.33
357,955
504,342
83,430
-1.17
3.26
% Change, over last 6 years
0.102
0.051
0.036
2.06
3.22
0.070
0.032
0.031
1.18
1.63
Employment-Population ratio
0.535
0.579
0.461
-1.41
3.49
0.602
0.569
0.467
0.64
3.63
Change, over last 6 years
0.041
0.047
0.023
-0.68
2.54
0.045
0.038
0.028
0.39
1.57
Manufacturing Labor Share
0.314
0.251
0.252
2.35
3.12
0.296
0.227
0.251
1.60
1.17
Change, over last 6 years
-0.014
-0.031
-0.008
1.52
-0.64
-0.030
-0.040
-0.007
0.87
-3.17
18.8
25.6
7.98
-1.35
3.02
2.75
3.92
2.38
-1.14
0.70
190,039
181,454
123,187
0.25
2.14
217,950
178,958
132,571
0.41
1.25
0.082
0.082
0.118
0.01
-0.97
-0.061
0.177
0.182
-1.23
-3.38
1,508
1,168
877
1.52
2.43
1,738
1,198
1,050
0.92
1.33
0.122
0.081
0.115
0.81
0.14
0.160
0.023
0.144
0.85
0.13
# of Counties
County Characteristics:
Total Per-capita Earnings ($)
% Change, over last 6 years
Population
# of Sample Plants
Plant Characteristics:
Output ($1000)
% Change, over last 6 years
Hours of labor (1000s)
% Change, over last 6 years
Winning
Counties
(1)
47
Losing
Counties
(2)
73
17,418
Within Same Industry (2-digit SIC)
Losing
All US
(6) -(7)
Counties
Counties
t-stat
(7)
(8)
(9)
19
(6)- (8)
t-stat
(10)
Notes: For each case to be weighted equally, counties are weighted by the inverse of their number per-case. Similarly, plants are weighted by the inverse of their
number per-county multiplied by the inverse of the number of counties per-case. The sample includes all plants reporting data in the ASM for each year between
the MDP opening and eight years prior. Excluded are all plants owned by the firm opening a MDP. Also excluded are all plants from two uncommon 2-digit SIC
values so that subsequently estimated clustered variance matrices would always be positive definite. The sample of all United States counties excludes winning
counties and counties with no manufacturing plant reporting data in the ASM for nine consecutive years. These other United States counties are given equal weight
within years and are weighted across years to represent the years of MDP openings. Reported t-statistics are calculated from standard errors clustered at the county
level. All monetary amounts are in 2006 US dollars.
180
Table 4. Incumbent Plant Productivity, Relative to the Year of a MDP Opening
Event Year
In Winning
In Losing
Counties
Counties
(1)
(2)
0.067
0.040
(0.058)
(0.053)
0.047
(0.044)
0.041
(0.036)
-0.003
0.028
(0.046)
0.021
-3
T= -2
T= 1
T =4
R-squared
Observations
(3)
0.012
(0.030)
-0.013
(0.022)
0.001
(0.011)
0
0.011
(0.022)
-0.003
(0.027)
0
-0.010
0.013
(0.018)
0.023
(0.026)
0.004
(0.036)
0.003
(0.047)
0.004
(0.053)
-0.023
(0.069)
"I"
0.027
(0.032)
0.018
(0.023)
0.020
(0.025)
-0.015
(0.024)
0.024
(0.021)
-0.005
(0.028)
0
(0.040)
(0.030)
t=
Difference
(1)-(2)
0.023
(0.019)
0.05 1*
(0.023)
0.050
(0.033)
0.076+
(0.043)
0.076*
(0.033)
0.077*
(0.011)
-0.028
(0.024)
-0.046
(0.046)
-0.073
(0.057)
-0.072
(0.062)
-0.100
(0.067)
(0.035)
0.9861
28732
Notes: Standard errors are clustered at the county level. Columns I and 2 report coefficients from the same
regression: the natural log of output is regressed on the natural log of inputs (all worker hours, building capital,
machinery capital, materials), year x 2-digit SIC fixed effects, plant fixed effects, case fixed effects, and the reported
dummy variables for whether the plant is in a winning or losing county in each year relative to the MDP opening.
When a plant is a winner or loser more than once, it receives a dummy variable for each incident. Plant-year
observations are weighted by the plant's total value of shipments eight years prior to the MDP opening. Data on
plants in all cases is only available 8 years prior to the MDP opening and 5 years after. Capital stocks were
calculated using the permanent inventory method from early book values and subsequent investment. The sample of
incumbent plants is the same as in columns 1 - 2 of Table 3. ** denotes significance at 1% level, * denotes
significance at 5% level, + denotes significance at 10% level.
181
Table 5. The Effect of the Opening of a MDP Plant on the Productivity of Incumbent Plants
All Counties
MDP Winners MDP Losers
(1)
(2)
MDP Counties
MDP Winners MDP Losers
(3)
(4)
All Counties
Random
Winners
(5)
Model 1:
Mean Shift
R-squared
Observations
(plant x year)
Model 2:
Effect after 5 years
Level Change
Trend Break
Pre-trend
0.0442+
(0.0233)
0.0435+
(0.0235)
0.0524*
(0.0225)
0.0477*
(0.0231)
[$170m]
-0.0824**
(0.0177)
0.9811
418064
0.9812
418064
0.9812
50842
0.9860
28732
0.9828
426853
0.1301*
(0.0533)
0.1324*
(0.0529)
0.1355**
(0.0477)
0.1203*
(0.0517)
[$429m]
-0.0559+
(0.0299)
0.0277
(0.0241)
0.0171+
(0.0091)
-0.0057
(0.0046)
0.0251
(0.0221)
0.0179*
(0.0088)
-0.0058
(0.0046)
0.0255
(0.0186)
0.0183*
(0.0078)
-0.0048
(0.0046)
0.0290
(0.0210)
0.0152+
(0.0079)
-0.0044
(0.0044)
-0.0197
(0.0312)
-0.0060
(0.0072)
-0.0057**
(0.0029)
0.9828
0.9861
0.9813
0.9812
0.9811
R-squared
426853
28732
50842
418064
418064
Observations
(plant x year)
YES
YES
YES
YES
YES
Plant & Ind-Year FEs
N/A
YES
YES
YES
NO
Case FEs
All
-7 <z 5
All
All
All
Years Included
Notes: The table reports results from the fitting of several versions of equation (8). Specifically, entries are from a regression
of the natural log of output on the natural log of inputs, year x 2-digit SIC fixed effects, plant fixed effects, and case fixed
effects. In Model 1, two additional dummy variables are included for whether the plant is in a winning county 7 to 1 years
before the MDP opening or 0 to 5 years after. The reported mean shift indicates the difference in these two coefficients, i.e.,
the average change in TFP following the opening. In Model 2, the same two dummy variables are included along with preand post-trend variables. The shift in level and trend are reported, along with the pre-trend and the total effect evaluated after
5 years. In columns (1), (2), and (5), the sample is composed of all manufacturing plants in the ASM that report data for 14
consecutive years, excluding all plants owned by the MDP firm. In these models, additional control variables are included
for the event years outside the range from r = -7 through T = 5 (i.e., -20 to -8 and 6 to 17). Column (2) adds the case fixed
effects that equal 1 during the period that r ranges from -7 through 5. In columns (3) and (4), the sample is restricted to
include only plants in counties that won or lost a MDP. This forces the industry-year fixed effects to be estimated solely
from plants in these counties. Incumbent plants are now required to be in the data only when the MDP opens and all 8 years
prior (not also for 14 consecutive years, though this does not change the results). For column (4), the sample is restricted
further to include only plant-year observations within the period of interest (where r ranges from -7 to 5). This forces the
industry-year fixed effects to be estimated solely on plant by year observations that identify the parameters of interest. In
column (5), a set of 47 plant openings in the entire country were randomly chosen from the ASM in the same years and
industries as the MDP openings. For all regressions, plant-year observations are weighted by the plant's total value of
shipments eight years prior to the opening. Plants not in a winning or losing county are weighted by their total value of
shipments in that year. All plants from two uncommon 2-digit SIC values were excluded so that estimated clustered
variance-covariance matrices would always be positive definite. In brackets is the value in 2006 US$ from the estimated
increase in productivity: the percent increase is multiplied by the total value of output for the affected incumbent plants in
the winning counties. Standard error clustered at the county level in parenthesis. ** denotes significance at 1% level, *
denotes significance at 5% level, + denotes significance at 10% level.
182
Table 6 The Effect of the Opening of a MDP Plant on the Productivity of Incumbent Plants. Robustness to Different
Specifications
Baseline
specification
Model 1: Mean Shift
Input industry
interactions
(3)
Inputwinner,
input-post
(4)
Region Year FE
(1)
Translog
functional
form
(2)
0.0477*
(0.0231)
0.0471*
(0.0226)
0.0406+
(0.0220)
0.0571*
(0.0245)
Model 2: After 5 years
(5)
Region Year Industry FE
(6)
Fixed Input
Cost Shares:
plant level
(7)
Fixed Input
Cost Shares:
SIC-3 level
(8)
0.0442+
(0.0230)
0.0369+
(0.0215)
0.0364
(0.0228)
0.0325
(0.0241)
0.1203*
0.1053+
0.0977*
0.1177*
0.1176*
0.0879*
0.0971
0.0938+
(0.0517)
(0.0535)
(0.0487)
(0.0538)
(0.0520)
(0.0442)
(0.0656)
(0.0538)
Notes: The table reports results from the fitting of several versions of equation (8). Column 1 reports estimates from the same specification as in column 4 of
Table 5. Column 2 uses a translog functional form. Column 3 allows the effect of each input to differ at the 2-digit SIC level. Column 4 allows for the effect
of inputs to differ in winning/losing counties and before/after the MDP opening. Column 5 controls for region (9 census divisions) by year fixed effects.
Column 6 includes census division by year by industry fixed effects. Column 7 reports estimates when fixing the coefficient on each input to be its average
cost share for each plant over the sample period. Per-period capital costs were calculated from capital rental rates using BLS data. Column 8 fixes the
coefficient to be the average across all plants in each 3-digit SIC. Standard error clustered at the county level in parenthesis. ** denotes significance at 1%
level, * denotes significance at 5% level, + denotes significance at 10% level.
183
Table 7. The Effect of the Opening of a MDP Plant on the Productivity of Incumbent Plants. Olley-Pakes (1996) and
Levinsohn-Petrin (2003) Specifications.
Baseline
specification
(1)
Model 1: Mean Shift
0.0477*
(0.0231)
Investment
(2)
0.0460*
(0.0230)
Investment,
Capital
(3)
0.0431+
(0.0226)
InvestmentCapital
Interactions
Materials
(4)
(5)
0.0399+
(0.0213)
0.0451+
(0.0230)
MaterialsCapital
Interactions
(6)
0.0399+
(0.0216)
Investment,
Materials,
Capital
(7)
0.0410+
(0.0222)
MaterialsCapital and
InvestmentCapital
Interactions
(8)
0.0397*
(0.0199)
0.0919+
0.1004*
0.1153*
0.1004*
0.0989*
0.0971*
0.1149*
0.1203*
(0.0487)
(0.0493)
(0.0487)
(0.0526)
(0.0490)
(0.0482)
(0.0517)
(0.0529)
Notes: The table reports results from the fitting of several versions of equation (8). Column (1) reports estimates from the same specification as in column 4 of
Table 5. To this baseline specification, the column (2) specification adds 4 th degree polynomial functions of log building investment and log
machinery/equipment investment. The column (3) specification adds 4 th degree polynomials in the two different capital stocks to the polynomials in
investment in the column (2) equation. The column (4) specification adds all the "own" interactions between polynomials in current investment and capital
(i.e., the building investment polynomial is interacted with the building capital stock polynomial, but is not also interacted with the machinery/equipment
polynomials for stocks or investment). Column (5) adds a 4 th degree function of log materials to the baseline specification. In the column (6) specification, a
4 th degree polynomial in materials is fully interacted with 4 th degree polynomials in building capital and machinery capital. Column (7) includes fourth-degree
polynomials in log materials and in log investment, and log capital stock for both types of capital (not interacted). Finally, column (8) includes the controls
from the columns (4) and (6) specifications. Standard error clustered at the county level in parenthesis. ** denotes significance at 1% level, * denotes
significance at 5% level, + denotes significance at 10% level.
Model 2: After 5 years
184
Table 8 The Effect of the Opening of a MDP Plant on the Productivity of Incumbent Plants, For Incumbent Plants in the
MDP's 2-digit Industry and All Other Industries
All Industries
Model 1:
Mean Shift
R-squared
Observations
Model 2:
Effect after 5 years
Level Change
Trend Break
Pre-trend
(1)
MDP's
2-digit Industry
(2)
All Other
2-Digit Industries
(3)
0.0477*
(0.0231)
[$170m]
0.1700*
(0.0743)
[$102m]
0.0326
(0.0253)
[$104m]
0.9860
28732
0.9861
28732
0.1203*
(0.0517)
[$429m]
0.3289
(0.2684)
[$197m]
0.0889+
(0.0504)
[$283m]
0.0290
(0.0210)
0.0152+
(0.0079)
-0.0044
(0.0044)
0.2814**
(0.0895)
0.0079
(0.0344)
-0.0174
(0.0265)
0.0004
(0.0171)
0.0147+
(0.0081)
-0.0026
(0.0036)
R-squared
0.9861
0.9862
Observations
28732
28732
Notes: The table reports results from the fitting of several versions of equation (8). For comparison, Column 1 reports the same results as column 4 of Table 5.
Columns 2 and 3 report estimates from the same regression, which fully interacts the winner/loser and pre/post variables with indicators for whether the
incumbent plant is in the same 2-digit industry as the MDP or a different industry. Standard error clustered at the county level in parenthesis. The numbers in
brackets are the value (2006 US$) from the estimated increase in productivity: the percent increase is multiplied by the total value of output for the affected
incumbent plants in the winning counties. ** denotes significance at 1% level, * denotes significance at 5% level, + denotes significance at 10% level.
185
Table 9. How the Effect of the Opening of a MDP Plant on the Productivity of Incumbent Plants Varies with Measures of
Economic Distance
CPS Worker Transitions
Citation pattern
Technology Input
Technology Output
(1)
0.0701 **
(0.0237)
(2)
(4)
(3)
(5)
(6)
0.0545**
(0.0192)
(7)
0.0374
(0.0260)
0.0256
(0.0208)
0.0320+
(0.0173)
0.0501
(0.0421)
0.0004
(0.0434)
0.0596**
(0.0216)
-0.0473
(0.0289)
0.0060
(0.0123)
Manufacturing Input
Manufacturing Output
0.0150
(0.0196)
-0.0145
(0.0230)
R-squared
0.9852
0.9852
0.9851
0.9852
0.9851
0.9852
0.9853
Observations
23397
23397
23397
23397
23397
23397
23397
Notes: The table reports results from the fitting of several versions of equation (8). Building on the Model I specification in column 4 of Table 5, each column
adds the reported interaction terms between winner/loser and pre/post status with the indicated measures of how an incumbent plant's industry is linked to its
associated MDP's industry. For assigning this linkage measure, the incumbent plant's industry is held fixed at its industry the year prior to the MDP opening.
Whenever a plant is a winner or loser more than once, it receives an additive dummy variable and interaction term for each occurrence. These industry linkage
measures are defined and described in Table 2 and here the measures are normalized to have a mean of zero and a standard deviation of one. The sample of
plants is that in column 4 of Table 5, but it is restricted to plants that have industry linkage data for each measure. Standard errors clustered at the county level are
reported in parentheses. ** denotes significance at 1% level, * denotes significance at 5% level, + denotes significance at 10% level.
186
Table 10. The Effect of the Opening of a MDP Plant on Wages and Number of Plants in the County
Panel 1
(Census of Manufacturers)
Dep. Var.: Log(Plants)
Dep. Var.:
Log(Total Output)
(1)
(2)
Difference-in-difference
0.1255*
(0.0550)
0.1454
(0.0900)
Panel 2
(Census of Population)
Dep. Var.: log(Wage)
(3)
0.0268+
(0.0139)
R-squared
0.9984
0.9931
0.3623
Observations
209
209
1057999
Notes: The table reports results from the fitting of three regressions. In Panel 1, the dependent variables are the log of number of establishments and the log of
total manufacturing output in the county, respectively. Controls include case effects, county, and year fixed effects. Entries are the county-level diff in diff
estimates for winning a MDP, based on data from the Census of Manufacturers. Because data is available every 5 years, depending on the Census year relative
to the MDP opening, the sample years are 1 - 5 years before the MDP opening and 4 - 8 years after the MDP opening. Thus, each MDP opening is associated
with one earlier date and one later date. The column (1) model is weighted by the number of plants in the county in years -6 to -10 and the column (2) model is
weighted by the county's total manufacturing output in years -6 to -10. In Panel 2, the dependent variable is log wage and controls include dummies for
age*year, age-squared*year, education*year, sex*race*Hispanic*citizen, and case fixed effects. The entry is the county-level diff in diff estimate for winning a
MDP. The pre-period is defined as the most recent census before the MDP opening. The post-period is defined as the most recent census 3 or more years after
the MDP opening. Thus, the sample years are 1 - 10 years before the MDP opening and 3 - 12 years after the MDP opening. The sample is limited to
individuals who worked last year, worked more than 26 weeks, usually work more than 20 hours per week, are not in school, are at work, and who work for
wages in the private sector. The number of observations reported refers to unique individuals - some IPUMS county groups include more than one FIPS, so all
individuals in a county group were matched to each potential FIPS. The same individual may then appear in more than one FIPS and observations are weighted
to give each unique individual the same weight (i.e., an individual appearing twice receives a weight of '/2).
Standard errors clustered at the county level are reported in parentheses. ** denotes significance at 1% level, * denotes significance at 5% level, + denotes
significance at 10% level.
187
Appendix Table 1. The Effect of the Opening of a MDP Plant on Inputs and Output (Unadjusted for Inputs) of Incumbent
Plants
Model 1: Mean Shift
(5)
Capital /
Labor
(6)
Energy /
Capital
(7)
0.0911**
(0.0302)
-0.0251
(0.0418)
0.0008
(0.0372)
Machinery
Capital
(3)
Building
Capital
(4)
Materials
(1)
Worker
Hours
(2)
0.1200**
(0.0354)
0.0789*
(0.0357)
0.0401
(0.0348)
0.1327+
(0.0691)
Output
-0.1310
0.0585
-0.0089
-0.0077
0.0509
0.0826+
0.0562
(0.1228)
(0.0493)
(0.0375)
(0.0541)
(0.0469)
(0.0300)
(0.0478)
Notes: Standard errors clustered at the county level are reported in parentheses. All outcome variables are run in logs. In columns (6) and (7), "Capital" is the
combined stock of building and machinery capital. In column (7), "energy" is the combined cost of electricity and fuels: 163 plant-year observations are
excluded that have missing or zero values for energy. See the text for more details. ** denotes significance at 1% level, * denotes significance at 5% level, +
denotes significance at 10% level.
Model 2: After 5 years
188